Contrast Enhancement In An Image

Abstract— Image enhancement is an important step in image processing. Among them, the contrast enhancement technique plays a vital role to bring out information that exists in low dynamic range of an image. Many images suffer from low contrast due to poor light conditions or due to devise itself. Contrast in an image makes object distinguishable from the background. It depends on luminance of the image. Currently a wide range of techniques exist for contrast enhancement. Some of them are better but time consuming while others are poor but fast, thus becoming selective in application. The aim of this paper is to provide information about them.

Index Terms— Image Enhancement, contrast, luminance.

Introduction

CONTRAST is the difference in brightness or luminance between objects or regions that makes them distinguishable. For applications of Digital images in everyday life, sufficient contrast is needed to make objects distinguishable. In human vision, contrast is defined by the change in color and brightness of object with respect to its surrounding. Due to various reasons like poor weather and light condition or due to device limitations, the image can suffer to low contrast. Hence, contrast enhancement techniques are needed to transform these images. One of the clear examples of Contrast enhancement is in medical imaging. Like, Enhancement of contrast in particular areas like chest radiography and mammography due to small difference in the x-ray attenuation coefficients.

Contrast Enhancement

Contrast Enhancement techniques can be divided into two parts. One is spatial domain method and other is frequency or transform based method. Spatial methods directly perform on the image pixel values. On the other hand, frequency-based methods manipulate image coefficients obtained by transforms such as Fourier or Wavelet. Linear contrast stretching and histogram equalization method are two main methods used in Spatial domain.

Linear Contrast Stretching

Contrast stretching or normalization tries to enhance the image by stretching the range of intensity value image contains to make full use of the 0 to 255 range. The technique is restricted to only linear mapping from input to output values. The result is less harsh, but tends to avoid the appearance of artificially created image. The equation used is [1, Eq.1]:

s=(r-c)((b-a)/(d-c))+ a (1)

Where a and b are the limits of the image (for standard 8-bit grayscale it is 0 to 255), c and d are value limits of original image and s and r are the new and original pixel value respectively.

Histogram Equalization

Histogram Equalization is one of the most used method for enhancing the contrast of image. It is simple and has relatively better performance on all images. It is performed by remapping of gray levels of the image based on statistical probability distribution of input grey levels. The probability is given by [1] Eq.2:

p_k= n_k/N ,k=0,1,…….,m-1 (2)

Where m is the maximum grey level, nk is the kth grey level and N is the total number of pixels. The cumulative density function for image is defined as:

c(k)= ∑_j^(k-1)▒〖p_(j ) (3)〗

Using (3) the transform function for the image is written as:

f(k)= X_(o )+(X_t-X_(o ) )c(k) (4)

Where X is discrete grey level. The final image is derived as:

Y=f(X(i,j)), Ɐ X(i,j) ͼ X (5)

The original image was enhanced using this method as shown in Fig.3.a.

(a) (b)

Despite for its fame, the Histogram Equalization is not used in consumer electronics such as television. As it is applied globally it is possible that there is infrequently loss of pixel intensity. To overcome this drawback advanced algorithms are used. Some of them proceed using regional histograms so that contrast can be enhanced in small regions. These methods are collectively known as Adaptive Histogram Equalization.

Adaptive Histogram Equalization

AHE is an excellent contrast enhancement technique used in both natural images and medical imaging. AHE is applied by using a sliding window that moves across the whole image. For every iteration, HE is generated around central pixel of the window (its contextual region). The cumulative grey level is calculated according to which ranks are decided. Thereafter each pixel is mapped to an intensity based on its rank.

The problem with AHE is that it is very slow. As well as it is noted that AHE suffers from excessive enhancement. This problem is solved by using another method known as Clipper Limited Adaptive Histogram Equalization (CLAHE). In this the local contrast gain is limited by restricting or clipping the height of histogram. It has better noise performance than AHE. The only limitation of this method is that it gives less contrast and appears more artificial.

Conclusion

Contrast enhancement plays an important role in current world of Digital Image Processing. Its applications are unlimited. In this paper, many techniques for contrast enhancement were discussed. However new techniques are researched and implemented each day. Each technique is unique and depends on the specific task, content image, type of viewer and cost of process.