Color correction - Where to begin ?

[budwzr]

Red? Jeez, my monitor is really off So I'm gonna have to excuse myself from discussing this topic.

[sanlyn]

Red? Jeez, my monitor is really off So I'm gonna have to excuse myself from discussing this topic.

Forbid! Use a pixel reader and look over some objects, like blacks, white, grays, etc. -- hues that are supposed to be neutral.

I tell ya, folks, that sample2.mkv color and blown-out brights are a hoot. I played with it most of the evening, but I guess I should shut up and post something that might be useful. Back in just a small while.

[smrpix]

Here's a before and after from resolve lite, legalised and neutrally balanced. Your originals don't really need a lot of work. Resolve Lite is free by the way, and may or may not import your DSLR files natively.

[budwzr]

Whew, is that the space shuttle caukpit?

P.S. I had to get around the auto-censor with "cauk" instead of "c*o*c*k".

[smrpix]

Whew, is that the space shuttle caukpit?

That's only the software version. When you add the 5-6 figure console it's way more impressive!

[sanlyn]

I don't know about other readers, but sample2.mkv looks re-encoded to me, but it's debatable. Despite the high bitrate it's rather noisy and looks ragged. The PCM audio sounds "normalized" and looks that way in the TMPGenc editor window below. Those little white peaks in the image indicate over-modulation distortion -- or the camera is recording out of spec audio. Sounds very scratchy.


[Attachment 23472 - Click to enlarge]

Nothing seems to get color that looks real or convincing. There are subtle color blotches that show up in skin tones, and you can faintly see the guy's complexion change a few time during the clip -- hard to see at first, until you watch it about 50 times. I even fed it to AfterEffects for white balance, and got a guy with skin the color of cantaloupe fruit. Add that to the noisy, dancing grain in the background, and I kept having the odd sensation that I was working with VHS.

I used AVisynth and Virtualdub. I used different coding than jagabo's earlier post, although jagabo got color and levels into the right neighborhood -- which ain't easy here. I made the following choices about color and said to hell with everything else: I think the microphone should be dark gray instead of dark gunmetal red, the guy's teeth shouldn't look as if he has bleeding gums, the t-shirt under his orange shirt probably isn't peach-colored or pink (well, you never know), his eyes probably aren't bloodshot, the elbow patch on his coat sleeve really wants to be black all the time instead of greenish brown or blue, his hair probably isn't blue, purple, or dark crimson, and his face probably isn't orange, or green. The sheets of paper in his hand have some sort of coloration; if you make those sheets into shades of pure grays or whites, everything else in the picture looks screwy. For tricky levels, try to keep the guy's facial highlights from blowing away into the ether. I'd rather the guy's coat didn't glow like a strobe light. Reduce red saturation too much and he turns green.

I decoded the sample mkv to lossless AVI with TMPGEnc Mastering Works, then I started with the script below. The "highlighter" function is from doom9. After trying several denoisers I ran a modification of 2BDecided's mod of an MVDegrain2 routine. I saved the script's output for Virtualdub later.

LoadPlugin("D:\AviSynth 2.5\plugins\plugins2\mt_masktools.dll")
LoadPlugin("D:\AviSynth 2.5\plugins\plugins2\masktools.dll")
LoadPlugin("D:\AviSynth 2.5\plugins\plugins2\mvtools.dll")

AviSource("Drive:\path\to\Sample2.avi")
ColorMatrix(mode="Rec.709->Rec.601") 
ColorYUV(cont_y=-20,off_y=-4,cont_v=-20)
SmoothTweak(saturation=0.95)
SmoothLevels(14, 0.90, 255, 16, 235,chroma=200,limiter=0,tvrange=true,dither=100,protect=6)
highlighter(threshold=25)

source=last #save original
backward_vec2 = source.MVAnalyse(isb = true, delta = 2, pel = 2, overlap=4, sharp=1, idx = 1)
backward_vec1 = source.MVAnalyse(isb = true, delta = 1, pel = 2, overlap=4, sharp=1, idx = 1)
forward_vec1 = source.MVAnalyse(isb = false, delta = 1, pel = 2, overlap=4, sharp=1, idx = 1)
forward_vec2 = source.MVAnalyse(isb = false, delta = 2, pel = 2, overlap=4, sharp=1, idx = 1)
source.MVDegrain2(backward_vec1,forward_vec1,backward_vec2,forward_vec2,thSAD=400,idx=1)

clean=last #save cleaned version
diff1=subtract(source,clean).Blur(0.25)
diff2=diff1.blur(1.5,0)
diff3=subtract(diff1,diff2) #diff3 is high-ish frequency noise only
sharpen(0.4,0.0) # sharpen cleaned version a little

#----- mix high frequency noise back in
overlay(last,diff3.levels(128,1.0,255,0,127,coring=false),mode="add", opacity=0.7)
overlay(last,diff3.levels(0,1.0,127,128,255,coring=false).Invert(),mode="subtract", opacity=0.7)
mergechroma(clean.aWarpSharp(depth=20.0, thresh=0.75, blurlevel=2, cm=1))
return last

Text of the "Highlighter" function that tries to impose some sanity onto wild highlights:

function Highlighter(clip v, float "gblur", bool "gradient", int "threshold", bool "twopass", int "amount", bool "softlimit", int "method")
{
gradient = default (gradient,true)
gblur = (gradient==true) ? default (gblur,100) : default (gblur,5.0) 
threshold = default (threshold,100) #The lower the value, the more sensitive the filter will be.
twopass = default (twopass,false)
amount = default (amount,10)
softlimit = default (softlimit,false)
method = default (method, 1)

amount = (amount<0) ? abs(amount) : amount

lighten=v.Tweak(sat=0).mt_lut("x "+string(threshold)+" > 0 x ?")
blurred= (gradient==true) ? lighten.gaussianblur(gblur).gaussianblur(gblur+100).gaussianblur(gblur+200) : lighten.gaussianblur(gblur)
fuzziness_mask=blurred.mt_edge(mode="prewitt", Y=3, U=2, V=2).mt_expand(mode="both", Y=3, U=2, V=2)
photoshop_overlay = (method==1) ? mt_lutxy(v,blurred.invert(),"x 127.5 > y 255 x - 127.5 / * x 255 x - - + y x 127.5 / * ? ") : \
                    (method==2) ? mt_lutxy(v,blurred.invert(),"255 255 x - 255 y - * 255 / -") : v.Tweak(bright=amount)
photoshop_overlay = (twopass==true) ? photoshop_overlay.SmoothLevels(gamma=2.0) : photoshop_overlay
merged=mt_merge(v,photoshop_overlay,blurred)
fuzzy= (softlimit==true) ? mt_merge(merged,mt_lutxy(v,merged,"x y + 2 /"),fuzziness_mask) : merged
return fuzzy
}

Things looked better after struggling with ColorMill and Gradation Curves for a few hours. AVisynth die-hards can load the VirtualDub filters and their settings into an Avisynth script, as I did here:

LoadVirtualDubPlugin("D:\VirtualDub\plugins\ColorMill.vdf", "ColorMill", 1)
LoadVirtualDubPlugin("D:\VirtualDub\plugins\gradation.vdf", "Curves1", 1)

AviSource("Drive:\path\to\Sample2_05D_2B.avi")
AddGrainC(2.0, 2.0)
ConvertToRGB32(matrix="Rec601",interlaced=false)

#------ primary color correction
ColorMill(25698, 24676, 26210, 24200, 26468, 28772, 26471, 25703, 25700, 25700, 25700, 25698, 25700, 1124, 6148)

Curves1(1,"000102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f202122232425262728292a2b2c2d2e2f303132333435"\
+"363738393a3b3c3d3e3f404142434445464748494a4b4c4d4e4f505152535455565758595a5b5c5d5e5f606162636465666768696a6b6c6d6e6f7"\
+"07172737475767778797a7b7c7d7e7f808182838485868788898a8b8c8d8e8f909192939495969798999a9b9c9d9e9fa0a1a2a3a4a5a6a7a8a9aaa"\
+"bacadaeafb0b1b2b3b4b5b6b7b8b9babbbcbdbebfc0c1c2c3c4c5c6c7c8c9cacbcccdcecfd0d1d2d3d4d5d6d7d8d9dadbdcdddedfe0e1e2e3e4e5e"\
+"6e7e8e9eaebecedeeeff0f1f2f3f4f5f6f7f8f9fafbfcfdfeff0708090a0b0c0d0e0f0f101112131415161718191a1b1c1d1e1f1f2021222324252"\
+"62728292a2b2c2d2e2f2f303132333435363738393a3b3c3d3e3f3f404142434445464748494a4b4c4d4e4f4f505152535455565758595a5b5c5d5"\
+"e5f5f606162636465666768696a6b6c6d6e6f6f707172737475767778797a7b7c7d7e7f7f808182838485868788898a8b8c8d8e8f8f90919293949"\
+"5969798999a9b9c9d9e9f9fa0a1a2a3a4a5a6a7a8a9aaabacadaeafafb0b1b2b3b4b5b6b7b8b9babbbcbdbebfbfc0c1c2c3c4c5c6c7c8c9cacbccc"\
+"dcecfcfd0d1d2d3d4d5d6d7d8d9dadbdcdddedfdfe0e1e2e3e4e5e6e7e8e9eaebecedeeefeff0f1f2f3f4f5f6f7050607090a0b0c0d0e101112131"\
+"4161718191a1b1d1e1f20212224252627282a2b2c2d2e2f31323334353738393a3b3c3e3f4041424445464748494b4c4d4e4f50525354555658595"\
+"a5b5c5d5f6061626365666768696a6c6d6e6f70717374757677797a7b7c7d7e8081828384868788898a8b8d8e8f90919394959697989a9b9c9d9e9"\
+"fa1a2a3a4a5a7a8a9aaabacaeafb0b1b2b4b5b6b7b8b9bbbcbdbebfc1c2c3c4c5c6c8c9cacbcccdcfd0d1d2d3d5d6d7d8d9dadcdddedfe0e2e3e4e"\
+"5e6e7e9eaebecedeef0f1f2f3f4f6f7f8f9fafbfdfefffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"\
+"fffffffffffff0607090a0b0d0e10111214151618191b1c1d1f202123242527282a2b2c2e2f303233353637393a3b3d3e3f414244454648494a4c4"\
+"d4f50515354555758595b5c5e5f606263646667696a6b6d6e6f717273757678797a7c7d7e80818384858788898b8c8d8f909293949697989a9b9c9"\
+"e9fa1a2a3a5a6a7a9aaacadaeb0b1b2b4b5b6b8b9bbbcbdbfc0c1c3c4c6c7c8cacbcccecfd0d2d3d5d6d7d9dadbdddee0e1e2e4e5e6e8e9eaecede"\
+"ff0f1f3f4f5f7f8fafbfcfefffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff"\
+"fffffffffffffffffffffffffffffffffffffffffffffffffffff000102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f2"\
+"02122232425262728292a2b2c2d2e2f303132333435363738393a3b3c3d3e3f404142434445464748494a4b4c4d4e4f505152535455565758595a5"\
+"b5c5d5e5f606162636465666768696a6b6c6d6e6f707172737475767778797a7b7c7d7e7f808182838485868788898a8b8c8d8e8f9091929394959"\
+"69798999a9b9c9d9e9fa0a1a2a3a4a5a6a7a8a9aaabacadaeafb0b1b2b3b4b5b6b7b8b9babbbcbdbebfc0c1c2c3c4c5c6c7c8c9cacbcccdcecfd0d"\
+"1d2d3d4d5d6d7d8d9dadbdcdddedfe0e1e2e3e4e5e6e7e8e9eaebecedeeeff0f1f2f3f4f5f6f7f8f9fafbfcfdfeff,152222202020202020000fff"\
+"f0007fff70005d4ff0006b6ff0000ffff")

#------ secondary correction

Curves1(1,"000102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f202122232425262728292a2b2c2d2e2f3031323334353"\
+"63738393a3b3c3d3e3f404142434445464748494a4b4c4d4e4f505152535455565758595a5b5c5d5e5f606162636465666768696a6b6c6d6e6f707"\
+"172737475767778797a7b7c7d7e7f808182838485868788898a8b8c8d8e8f909192939495969798999a9b9c9d9e9fa0a1a2a3a4a5a6a7a8a9aaaba"\
+"cadaeafb0b1b2b3b4b5b6b7b8b9babbbcbdbebfc0c1c2c3c4c5c6c7c8c9cacbcccdcecfd0d1d2d3d4d5d6d7d8d9dadbdcdddedfe0e1e2e3e4e5e6e"\
+"7e8e9eaebecedeeeff0f1f2f3f4f5f6f7f8f9fafbfcfdfeff000102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f20212"\
+"2232425262728292a2b2c2d2e2f303132333435363738393a3b3c3d3e3f404142434445464748494a4b4c4d4e4f505152535455565758595a5b5c5"\
+"d5e5f606162636465666768696a6b6c6d6e6f707172737475767778797a7b7c7d7e7f808182838485868788898a8b8c8d8e8f90919293949596979"\
+"8999a9b9c9d9e9fa0a1a2a3a4a5a6a7a8a9aaabacadaeafb0b1b2b3b4b5b6b7b8b9babbbcbdbebfc0c1c2c3c4c5c6c7c8c9cacbcccdcecfd0d1d2d"\
+"3d4d5d6d7d8d9dadbdcdddedfe0e1e2e3e4e5e6e7e8e9eaebecedeeeff0f1f2f3f4f5f6f7f8f9fafbfcfdfeff000102030405060708090a0b0c0d0"\
+"e0f101112131415161718191a1b1c1d1e1f202122232425262728292a2b2c2d2e2f303132333435363738393a3b3c3d3e3f4041424344454647484"\
+"94a4b4c4d4e4f505152535455565758595a5b5c5d5e5f606162636465666768696a6b6c6d6e6f707172737475767778797a7b7c7d7e7f808182838"\
+"485868788898a8a8b8c8d8e8f90919293949596979898999a9b9c9d9e9fa0a1a2a2a3a4a5a6a7a8a9aaababacadaeafb0b1b2b2b3b4b5b6b7b8b9b"\
+"ababbbcbdbebfc0c0c1c2c3c4c5c6c6c7c8c9cacbcccccdcecfd0d1d2d2d3d4d5d6d7d8d8d9dadbdcdddddedfe0e1e2e3e3e4e5e6e7e8e8e9eaebe"\
+"cededeeeff0000102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f202122232425262728292a2b2c2d2e2f30313233343"\
+"5363738393a3b3c3d3e3f404142434445464748494a4b4c4d4e4f5052535455565758595a5b5c5d5e5f606162636465666768696a6b6c6d6e6f707"\
+"172737475767778797a7b7c7d7e7f808182838485868788898a8a8b8c8d8e8f90919293949595969798999a9b9b9c9d9e9fa0a1a1a2a3a4a5a5a6a"\
+"7a8a9a9aaabacacadaeafafb0b1b2b2b3b4b4b5b6b7b7b8b9b9babbbbbcbdbdbebfbfc0c1c1c2c3c3c4c5c5c6c6c7c8c8c9cacacbcbcccdcdcecfc"\
+"fd0d0d1d2d2d3d3d4d5d5d6d6d7d8d8d9d9dadadbdcdcddddde000102030405060708090a0b0c0d0e0f101112131415161718191a1b1c1d1e1f202"\
+"122232425262728292a2b2c2d2e2f303132333435363738393a3b3c3d3e3f404142434445464748494a4b4c4d4e4f505152535455565758595a5b5"\
+"c5d5e5f606162636465666768696a6b6c6d6e6f707172737475767778797a7b7c7d7e7f808182838485868788898a8b8c8d8e8f909192939495969"\
+"798999a9b9c9d9e9fa0a1a2a3a4a5a6a7a8a9aaabacadaeafb0b1b2b3b4b5b6b7b8b9babbbcbdbebfc0c1c2c3c4c5c6c7c8c9cacbcccdcecfd0d1d"\
+"2d3d4d5d6d7d8d9dadbdcdddedfe0e1e2e3e4e5e6e7e8e9eaebecedeeeff0f1f2f3f4f5f6f7f8f9fafbfcfdfeff,152222202020405020000ffff0"\
+"007fff70005d4ff0006b6ff0000ffff")

ConvertToYV12(interlaced=false)
colorMatrix(mode="Rec.601->Rec.709") 
return last

[Attachment 23473 - Click to enlarge]

[sanlyn]

It's codec "mushiness". It's a term people use to describe how the picture is soft & impaired from the poor compression . It's typical of poor AVC compression, and tends to affect shadow areas disproportinately. It's sort of a glazed over, low detail look. It's especially bad for early DSLR's , because they use baseline AVC (no cabac, no b-frames) - the bitrates are decently large 40-50Mb/s that you 'd expect a sharper picture - but still looks not that great for that bitrate - it falls apart when anything taxes the compression . Other cameras at 1/2 that bitrate produce cleaner/sharper pictures (of course there are dozens and dozens of other factors contributing, but a large difference is the compression)......
.........

Most consumer level Canon and Nikon's share the same recording module (and they record using crappy baseline AVC, very mushy picture), and they all tag it the same way. Those VUI flags are just suggestions as to how to view the YUV data in RGB (ie. suggestions to the player how to convert to RGB for display).....
.......

No wonder I kept thinking I was working with VHS again. My eyes kept saying "these problems look very familiar", but my brain said "no way".

Important thing you should be aware of - the choice decoder plays a large role in what levels you get for the native MOV quicktime files......
.....
Vegas in 8bit mode will use studio RGB (that's not the same thing as sRGB) - so superbrights/darks will be preserved. Studio RGB is almost the same thing as "full range" or a PC matrix if you're using avisynth. It's tough to do this properly in the studio version because you have no scopes - you're working half blind.

People kept telling me it was just my imagination and made "no difference".

Two most informative posts. Thanks.

[JDługosz]

This is not the same camera, but it's just a typical comparison of ML raw vs. native h.264 "mush" (and the h.264 isn't that mushy here, it's a lot worse when there is motion)

My video has not been that bad. I see from the link that it's a "low light" concern in particular. I'm improved things by using a f/2.8 zoom lens, rather than a f/4—f/5.6 depending on zoom. What I have noticed is scintillation of the plain background, which is darker than the lighted performers. I improved that with post-processing in Avidemux (I forget the filter name). I'll have to try things to see just what the limits are, like I've done with regular photos. Looking back, I do see that a dimmer stage gave bad results (even with the new lens) so the home stage lighting must be near the limit of its capabilities.

And if you're not going the raw route (maybe you have write speed or reliablity concerns), there are lots of free ways to do basic color correction with monitoring in avisynth, vdub, (avidemux doesn't have waveform/histogram etc... monitoring) on the native recordings

It didn't in the release version, but I see the 60D is listed now, so I'd have to go with fresh builds. I will try it now, but don't know if that will be an option in general.

That's avisynth for "auto" white balance.

I've not gotten avisynth to work. I found the idea of command-driven processing to be attractive though.

[JDługosz]

Here's a before and after from resolve lite, legalised and neutrally balanced. Your originals don't really need a lot of work. Resolve Lite is free by the way, and may or may not import your DSLR files natively.

Awesome! A free tool to give useful color-correction meters and such, that exactly fits the bill.
I'll download it momentarily.
Can you tell me what settings you ended up with?

[JDługosz]

Whew, is that the space shuttle caukpit?

That's what people call my desk. The three large monitors in a corner + laptop probably is what gets the attention first. The shock of the blank keyboards comes with closer inspection.

[smrpix]

Can you tell me what settings you ended up with?

All the settings were deliberately left visible in the images I posted. The tools are designed to be manipulated intuitively.

[JDługosz]

I don't know about other readers, but sample2.mkv looks re-encoded to me, but it's debatable.

That's copied by Avidemux without re-encoding.

The PCM audio sounds "normalized" and looks that way in the TMPGenc editor window below. Those little white peaks in the image indicate over-modulation distortion -- or the camera is recording out of spec audio. Sounds very scratchy, and normalization often makes certain sounds "blast" and fluctuate.

It sounded that way in person, too! I wish I had marshmallow earplugs, and will add them to my kit next time
Two large speakers were set up at a restaurant, and I was between them and only a couple feed downwind of them. The area used for the performances had a higher ceiling and it dropped lower over the bulk of the seating.

It was set to auto-level the sound, and I couldn't find the controls to turn down the sensitivity. Sound is another story. Next year I'll take audio from their mixing console to my 96khz raw wav recorder.

[poisondeathray]

If you're interested in resolve, there is a free 11 part tutorial series on resolve at aetuts+, but it probably covers more than the basics that you were initially looking for.
http://cgi.tutsplus.com/tutorials/davinci-resolve-fundamentals-introduction--ae-25208

Everyone has their own "process" as to how to approach color correction. But I would say most people deal with levels first - setting the black and white points (e.g. with levels or curve filters), then move onto other areas

avisynth, vdub , blender are free and have waveform, histograms , monitoring, color manipulation tools too. With a programming background, it should be easy to pick up avisynth (all script driven).

Those "scintillating backgrounds" are noise. If you want to get into cleaning up the noise, avisynth is helpful too . It's a different topic than color correction but there are many ways to approach that . It's slightly different than processing photos, because there is a temporal element to it (and temporal filters)

[JDługosz]

Can you tell me what settings you ended up with?

All the settings were deliberately left visible in the images I posted. The tools are designed to be manipulated intuitively.

If you mean the text under the color wheels, I can't read it. Blowing up just gives smudges.

[smrpix]

The adjustments on the wheels are very minor, use the positions of the indicators to get you close. The biggest changes are in the RGB curve controls which balanced the color -- again, try to match the visual clues, don't worry so much about numbers.

Is this any clearer (I tossed the full resolution version because it was a big file and I thought it wouldn't be needed again.)

[budwzr]

In this waveform I see clipping, not normalization.

[sanlyn]

Then the camera's clipping the audio. Maybe there's an onboard adjustment?

[JDługosz]

The adjustments on the wheels are very minor, use the positions of the indicators to get you close. The biggest changes are in the RGB curve controls which balanced the color -- again, try to match the visual clues, don't worry so much about numbers.

Is this any clearer (I tossed the full resolution version because it was a big file and I thought it wouldn't be needed again.)

[Attachment 23499 - Click to enlarge]

Thanks.

I tried my hand at Resolve, and I did much better than anything previously even without experience. So it is indeed a great tool for this.

But... can you tell me how to save as uncompressed? I looked through the Deliver options and found AVI/YUV-422, but then Avidemux can't load it. Which of the available choices would be uncompressed and give the best results when fed into another program to apply noise filters and encode using x264?

[JDługosz]

Then the camera's clipping the audio. Maybe there's an onboard adjustment?

ML gives audio level bars, and I noticed they were going into the red region but not hitting the top, which leaves a sticky shadow. So I figured the auto-level was working OK. Owell. Maybe it was picking up already clipped sound from the amplification.

[smrpix]

But... can you tell me how to save as uncompressed?

You'll have to experiment with that one. Almost every output option is a higher datarate/quality setting than what you're starting with. Avidemux can load at least some of the quicktimes.

[budwzr]

Then the camera's clipping the audio. Maybe there's an onboard adjustment?

ML gives audio level bars, and I noticed they were going into the red region but not hitting the top, which leaves a sticky shadow. So I figured the auto-level was working OK. Owell. Maybe it was picking up already clipped sound from the amplification.

The mic gain is probably too high, and no compressor or limiter available, so it clips.