Article Preview
Top1. Introduction
Image contrast enhancement is one of the main concerns in the arena of digital image processing. Its main objective is to produce better image quality for correct interpretation.
Study of literature reveals, histogram equalization (HE) is most widely used and popular image contrast enhancement technique, due to its less computation cost and easy implementation. It stretches dynamic range of gray levels, flattens the cumulative density function and successfully achieves overall image contrast enhancement (Gonzalez, & Woods, 2002). But the problem of intensity saturation leads to appearance of undesirable visual artefacts in a processed image. Thus, HE is inefficient in preserving the brightness and providing detailed information recorded in an enhanced image.
To overcome the above said limitations, researcher Kim in 1997 proposed a multiple segmentation-based approach (Kim, 1997). Kim applies Brightness Preserving Bi-Histogram Equalization (BBHE) technique for contrast enhancement of an input image. BBHE completes its task into two steps. At first step, it partitions an input image histogram into two sub histograms on the basis of its mean gray level. At second step, it equalizes each sub histogram independently (Kim, 1997). In this technique Kim proved that the mean brightness of processed image lies between mean brightness of an input image and middle gray level. This technique also reduces the annoying artefacts in the processed image.
Researchers in 1999 proposed another new technique namely Dualistic Sub Image Histogram Equalization (DSIHE). This technique is similar to BBHE except that it uses median rather than mean value. Experimental results in a simulation environment prove that DSIHE is better than BBHE in terms of preservation of average information content of an image (Chen, & Zhang, 1999).
Both techniques are effective in contrast enhancement. BBHE as well as DSIHE faces the problem of brightness shift and intensity saturation artefacts in a processed image.
Researchers Chen et al. in 2003 proposed Recursive Mean Separate Histogram Equalization (RMSHE), an extension of BBHE. In RMSHE, authors suggested, the recursive segmentation of an input image histogram on the basis of local mean. At each step, existing sub-histogram is partitioned into two sub histograms. After 
step, the algorithm yields, 
sub histograms, where r is a natural number. The number of recursive steps, r depends upon user’s choice. Moreover, the authors claimed that with increase in r, brightness of processed image becomes equal to brightness of an original input image (Chen, & Ramli, 2003). This technique preserves more brightness than BBHE. But it faces the problem of over enhancement at low contrast regions. In addition, it is challenging to find the best value of r.
Researchers Sim et al. in 2007 proposed Recursive Sub Image Histogram Equalization (RSIHE), an extension of DSIHE. This technique is similar to RMSHE except that it does the partitioning of histogram on the basis of its median (Sim, Tso, & Tan, 2007). The authors claimed that, this technique preserves more brightness and information content than DSIHE.