Photoshop v 4.0.1 Bugs Page

Troublesome bugs in Adobe Photoshop, version 4.0.1.

Monitor Setup has Error

The gamma strip in the calibration dialog has a 1 bit chessboard patter. Such dithering does not provide accurate reference level on any CRT monitor. This is quite an unbelievable bug in a pro level, market leader application. The error can be very large and is always notable, it depends on the quality of the monitor. More info.

Middle input box in the Levels dialog is not a gamma control

The User Guide refers to the middle Input Levels box in the Levels dialog as a Gamma control. This is not correct. It follows the gamma function closely at highlights but in shadows there is a serious Adobe specific tweaking. Result: shadows will not convert as they should. If this control is used for increasing the overall lightness of the image then the shadows will remain much too dark. Fro the same reason this control can not be used for applying a gamma compensation. More info.

Histogram in Levels dialog is inaccurate

The image was the testpic.jpg from /photoshop/calibrat directory. The error depends on the image content and the Zoom value. The same inaccuracy exists with the Image/Histogram dialog also.

This problem can be avoided by un-checking the "Use Cache for Histograms" in the File/Preference/Memory&ImageCache -dialog.

Actions do not play recorded Curves adjustment that use a saved curve

This is a very annoying bug, however not the only bug in the Actions. Why it is annoying? The Curves dialog is the only way to apply an input linearization. Because if this bug input linearization can not be automated, instead the input compensation needs to be applied each time for each new image manually (in case the input device is non-linear).

Color error in CMYK and Lab* modes

There is large errors in the Photoshop CMYK and L*a*b* color mode conversions. First a quick look at the chapter "About CMYK conversion", on page 100 in the User Guide. It says the following:

About CMYK conversion. Adobe Photoshop uses the Lab color mode when converting color values from one mode to another. Because Lab provides a system for defining color values in all modes, using Lab as an intermediate mode for color conversion ensures that the colors are not altered in the conversion process (other than the necessary clipping of out-of-gamut colors).

There is no need for clipping out out-of-gamut colors when doing an RGB to L*a*b conversion since the L*a*b gamut is much larger than the RGB gamut. This clipping is done only when converting to CMYK gamut.

Yet RGB to L*a*b converison it will reduce the amout of colors by 8 times.

Firstly we see that the conversion from RGB to CMYK goes via the L*a*b* space. So any errors in RGB to L*a*b* will affect to the printing quality.

The number of discrete colors available in the RGB space is 256*256*256 = 16,777,216.

In the L*a*b* mode of Photoshop the L* can have *integer* values from 0 to 100 so there are 101 discrete values. The a* and b* both can have integer values from -120 to +120 so they both can have one out of 241 discrete values. The result from this is that in the Photoshop L*a*b* mode there are only 101*241*241 = 5866181 discrete colors available. But this is not all. The L*a*b* space boundary (gamut) is larger than the RGB space. When an RGB image is converted into L*a*b* space the L* will be utilized fully (0 to 100) but the a* will only have values from 99 to -90 and b* can have values from 92 to -107. So this results that the 16,777,216 RGB values will be mapped into only 100*190*200=3,800,000 discrete colors in the Photoshop L*a*b* space.

The main problem is that this mapping is not done evenly (since the L*a*b mode is highly non-linear). The result is that some of the colors will change very much, some only a little and some remain intact. And there can be large difference in the error between nearby colors, such changes will be easily noticed.

How to minimize the Color mode errors

The largest error when going from RGB to Lab and back to RGB is +/- 21 levels (this error is in one of the primaries, other two can change also. That will increase the overall error still, I have not mapped it this way yet). This is the case when the value in the "Monitor Gamma" input box is 1.8. The error increases still if the "Monitor Gamma" value is increased.

If the "Monitor Gamma" value is set to 1.0 then the largest error is 'only' +/- 6 levels. This difference is due to the way Photoshop manages the "Monitor Gamma" setting. It applies an 1/"Monitor Gamma" compensation to the image data values. This generates quantization error and the error is then further amplified by the RGB to Lab conversion.

Also when the "Monitor Gamma" value is set to 1.0 the error is distributed quite evenly over the color space.

With 1.8 in the "Monitor Gamma" input box the error is concentrated to highlights. This is direct consequence due to the quantization caused by the "Monitor Gamma", it is highest in the highlights. And in addition in this setup the error concentrates also in saturation edges that are 120 degree away from that color plane that the error is inspected for (I do not understand the reason for this).

The ambient Light setting affects also to the conversion. Smallest average error is achieved when it is set to High. Smallest absolute error is achieved when it is set to Medium. The White Point and Phosphors -settings do not have much effect to the error in general.