Reading Histograms-- Part II

Article and Photography by Ron Bigelow

www.ronbigelow.com

Photoshop CS2 Used in this Tutorial

Figure 1: Leaf Image
Figure 1 shows an underexposed image of a leaf. The image appears dark, and the colors are muted.
Figure 2: Histogram of Leaf Image
The histogram of the underexposed image is shown in Figure 2. This histogram takes a bit of analysis. The histogram appears shifted to the left. This is due to two factors. The first factor is that the image is composed of a lot of darker colors. The second factor is that the image is underexposed. There is also some clipping of the dark tones on the left side of the histogram. Since this image was shot in JPEG, those tones are lost and can not be retrieved (if the image had been shot in raw, it might have been possible to retrieve the tones). The key point in this image is the large, empty area on the right side of the histogram. This indicates that the image contains few light tones. In addition, there are only a few midtones. By using a Curves adjustment to set the highlight values and to lighten the midtones, this image can be improved.
Figure 3: Curves Used to Adjust the Tonality
Figure 3 shows the rather strong Curves adjustment that was used to adjust the tonality of the image.
Figures 4 -- 5 show the leaf image before and after the Curves adjustment was made. Figures 6 -- 7 show the associated histograms. The difference between the before and after Histograms is small. The important thing to notice is that the small bulge in the middle of the before histogram, which represents the midtones, has been shifted to the right in the after Histogram. Thus, the large empty space on the right side of the Histogram has been reduced (setting the highlights) and the midtones have been shifted to the right (lightening the midtones).
Figure 4: Leaf Image before Tonal Adjustment
Figure 5: Leaf Image after Tonal Adjustment
Figure 6: Histogram before Tonal Adjustment
Figure 7: Histogram after Tonal Adjustment

While Photoshop was able to improve this image, one thing that should be noted is that using Photoshop to correct poor quality images is not a good practice. Images should be shot properly in the first place. Repairing poor images in the computer usually results in an image that is less than high quality. For instance, this image will have increased noise in the darker tones.

Figure 8: Mountain Sunset Image

Figure 8 shows a sunset over a mountain. This image is composed primarily of very dark tones and midtones. Light tones are mostly missing as would be expected of this type of shot.

Figure 9: Histogram of Mountain Sunset Image

The Histogram in Figure 9 shows the predominance of the dark and midtones. The Histogram verifies that the lighter tones are almost completely missing. Despite the fact that this histogram is missing the lighter tones, this image does not need any tonal editing as the histogram of this image shows the tones that would be expected based on the image content.

The Leaf and Mountain Sunset images again show the importance of comparing a Histogram to its image. The Histograms in both Figures 2 and 9 are lacking lighter tones. However, comparing Figure 2 to its image indicates an underexposed image while comparing Figure 9 to its image indicates a properly exposed image that is composed of mostly dark and midtones.

Figure 10: Canyon Image
Figure 10 shows a shot of a canyon. The image appears overexposed. More importantly, the fog in the upper right hand corner of the image appears to have little or no detail.
Figure 11: Histogram of Canyon Image

The image's Histogram in Figure 11 is shifted to the right, thus, verifying the overexposure. The Histogram also contains no dark tones. These problems can be improved with a tonal adjustment. Of far greater concern is the right hand side of the Histogram. It shows that the highlights were clipped (i.e., cut off). If the image is a JPEG (which this image is), the clipped tones are completely lost and can not be recovered. If the image was from a raw file, the clipped tones may be recoverable, but there is no guarantee that they can be recovered.

The loss of the highlights is shown in the close-up in Figure 12.

Figure 12: Loss of Clipped Highlights
Figure 13: Curves Used to Adjust the Tonality
Figure 13 shows the Curves adjustment that was used to adjust the tonality of the image.
Figures 14 -- 15 show the clipped image before and after the Curves adjustment was made. Figures 16 -- 17 show the associated histograms. The important thing to notice is that, although the overexposure problem was compensated for with the Curves adjustment, the clipped highlights are still clipped. Curves can not remedy this problem.
Figure 14: Canyon Image before Tonal Adjustment
Figure 15: Canyon Image after Tonal Adjustment
Figure 16: Histogram before Tonal Adjustment
Figure 17: Histogram after Tonal Adjustment
Figure 18: Canyon Sunset Image
Figure 18 shows a shot of a sunset over a canyon. There are a couple of problems with this image. First, the image lacks contrast. As a result, the sunset looks rather lifeless. Second, the mountains in the foreground lack detail; they appear to be completely black.
Figure 19: Histogram of Canyon Sunset Image
The Histogram of the image (see Figure 19) tells the story. The left side of the Histogram shows that the darker tones have been seriously clipped. As in the previous image, if the image is a JPEG, the clipped tones are completely lost and can not be recovered. Now, this image was shot in raw. However, some experimentation with the raw file showed that the dark tones were clipped so much that they are not recoverable even from the raw file. Thus, the darker tones are permanently lost.
Figure 20: Curves Used to Adjust the Tonality
Some life can be added to the sunset by increasing the contrast. However, this adjustment will do nothing for the lost tones. Figure 20 shows the Curves adjustment that was used to adjust the tonality of the image.
Figures 21 -- 22 show the clipped image before and after the Curves adjustment was made. Figures 23 -- 24 show the associated histograms. As can be seen, the sunset now has more contrast; however, the Curves adjustment has done nothing for the lost dark tones.
Figure 21: Canyon Sunset Image before Tonal Adjustment
Figure 22: Canyon Sunset Image after Tonal Adjustment
Figure 23: Histogram before Tonal Adjustment
Figure 24: Histogram after Tonal Adjustment

Channel Histograms

The grayscale Histogram does not tell the entire story. It essentially takes the information from the three color channels (i.e., the red, green, and blue channels) and calculates a grayscale tonality. This might be thought of as an average tonality of the three color channels. However, the three channels can be significantly different from each other. To understand the full tonality of the image, a photographer needs to look at each of the three channels.

Figure 25: Flower Image
Figure 25 shows a shot of a desert flower. This image could use a little tonal adjustment. This image also shows the danger of depending on only the grayscale Histogram. The image will be adjusted first with the grayscale Histogram. Then, the results will be analyzed with the Histograms of all three channels.
Figure 26: Histogram of Flower Image
Figure 26 shows the grayscale Histogram of the flower image. It appears that setting the highlight values would be a good first adjustment to improve the impact of this image. A contrast adjustment would also help. So, Levels is launched to make the adjustment.
Figure 27: Levels Used Adjust the Tonality
Figure 27 shows the Levels adjustment used to set the highlight values and the contrast. The White Input Slider was moved to the left to the point where the grayscale Histogram indicates that the significant detail starts. In addition, the Gamma Input Slider was moved to the left to adjust the contrast.
At first, this might appear to be a smart move. After all, Figures 28 -- 29 show that the image after the Levels adjustment looks better than the original image.
Figure 28: Flower Image before Tonal Adjustment
Figure 29: Flower Image after Tonal Adjustment
Figure 30: Channel Histograms before Levels Adjustment

However, a look at the channel Histograms tells a different story. Figure 30 shows the channel Histograms before the tonal adjustment. The important point for this tonal adjustment is that the red Histogram extends almost all of the way to the right side of the chart. In other words, setting the highlight values too far to the left will clip this channel and cause a loss of detail. In fact, if one looks at the channel Histograms after the tonal adjustment (see Figure 31), this is exactly what happened when the Levels adjustment was made. The red Histogram in Figure 31 shows that the red channel was clipped. This can be seen by the red spike at the right side of the red Histogram.

Figure 31: Channel Histograms after Levels Adjustment

 

Close-ups of the yellow flower in Figures 32 -33 verify this loss of detail. Figure 32 shows the flower before the Levels adjustment. Figure 33 shows the flower after the adjustment. Comparison of the two images clearly shows a loss of detail after the Levels adjustment.
Figure 32: Close-up of the Flower Image before Tonal Adjustment
Figure 33: Close-up of the Flower Image after Tonal Adjustment
Figure 34: Curves to Adjust the Tones without Clipping the Red Channel
A better approach would have been to consider the Histograms from all three channels before deciding how to edit the image. Considering this information, the image was reedited using the Curves adjustment in Figure 34. This adjustment protected the red channel from being clipped. This is shown in Figure 35, which shows the channel Histograms after the Curves edit. It can be seen that the red channel has not been clipped with this edit.
Figure 35: Channel Histograms after Curves Edit
Comparing the close-ups of the flower with the Levels adjustment (see Figure 36) and the flower with the Curves adjustment (see Figure 37) shows the superiority of the Curves adjustment. The Curves adjustment protected the detail in the red channel and resulted in a better image. The final image is shown in Figure 38.
Figure 36: Close-up of the Flower Image with Levels Adjustment
Figure 37: Close-up of the Flower Image with Curves Adjustment
Figure 38: Final Image

Channel Histogram Analysis

Analyzing the individual channel Histograms is similar to analyzing the grayscale Histogram. The main difference is that all three channel histograms need to be considered before making any editing choices. Subsequently, no further examples of channel Histograms will be presented.

Conclusion

Histograms are a critical tool for analyzing images and determining the results of edits. The ability to read histograms is a necessary skill for anyone who desires to perform digital image editing.

Articles

Histogram -- I