Performance Evaluation of Global Histogram Equalization technique using Neighbourhood Metrics over Facial Databases.

Contrast enhancement of facial images improves the recognisability of faces in the scene. Face images acquired in real time are low contrast images. Contrast of an image is clearly revealed in its histogram representation. Global Histogram equalization is a well-known contrast enhancement technique but it does not adapt to local information in the image. Localizing the global technique results in increasing the histogram spread. To increase the local contrast several neighbourhood metrics were proposed for use in conjunction with Histogram equalization. This paper investigates the use of competing neighbourhood metrics used with histogram equalization on facial databases. The experimental results bring out the comparative analysis and efficiency of how the various neighbourhood metrics sorts the pixels of same intensity into different sub-bins which increases the local contrast of an image. It is clear from the results that there are many large bins in the resultant image after equalizing the illumination affected image using Global histogram equalization with voting metric and distinction metric. Also the resultant image histogram shows that many bins are empty and annoying artifacts are still present in the resultant image.