Restoration of Old Color Photographs

AUTOMATIC WHITE BALANCING (GRAY WORLD ALGORITHM)

i. Get the average value of each color channel and save them in a variable, e.g. Rave, Gave, Bave respectively.

ii. Divide each RGB channel by its respective average value.

iii. Display the image.

The first image below is your original image, and the image on the right is your image after applying automatic white balancing.

REFERENCE WHITE BALANCING

i. Select a ‘region of interest’ in the image that is known to be white (e.g. a white shirt, white wall, white paper). Get the average value of each color channel in the ROI and save them, e.g. Rw, Gw, Bw respectively.

ii. Divide each RGB channel by the respective average value from the ROI.

iii. Display the image.

After comparing both of the resulting images from the automatic and reference white balancing. They actually look somehow similar for me. If the goal of the enhancement of the image is to make it more appealing, then both methods work effectively for me. But if I were to choose one of the two, I would choose the output from the automatic white balancing since it produced a better luster on the cars in the image.

Performing histogram manipulation on your image, which will improve its contrast.

i. Convert your RGB image to HSV using an appropriate function in your language of choice (e.g. color.rgb2hsv() in scikit-image, rgb2hsv() in MSO).

ii. Perform the histogram manipulation procedure on the value channel only of your HSV image (usually the last channel, consult documentation for details). You may choose your own target CDF.

iii. Reconstruct your adjusted HSV image by taking the original H and S channels with your histogram back projected V channel, and convert back to RGB.

iv. Display the reconstructed RGB image.

4. Comment on why the value channel was used in histogram back projection. How does HSV differ from RGB, and what do each of the color channels in HSV correspond to? What happens when we perform histogram backprojection on each RGB channel individually and display the RGB image?

The first image above is the original image to which histogram back projection is applied. The result after back projection on the HSV channels and RGB are shown above. HSV corresponds to hue, saturation and value. The hue component represents the color, the saturation represents the purity of the color while the value represents the intensity. Histogram back projection was only used on the value channel since it controls only the intensity and thus will preserve the image’s color. When applying histogram back projection in the RGB color space, and since these channels are correlated to each other then there is a possibility that the original color will not be preserved unlike with the HSV color space where only the hue component represents the color.