Enhancing an image is the highly demanded digital image processing task, as all the other piece of work depend on how well the image is conveying the information. The core purpose of enhancing an image is in identifying the information details hidden in an image. Image Enhancement is necessary in improvising the quality of image for human perception. Contrast adjustment, noise removal from an image, highlighting of an image detail are few operations of image enhancement. Enhancing an image can be broadly done in two ways by considering only the pixel of interest or by considering the neighbouring pixels of the interested pixel i.e spatial domain or by converting the image into Fourier transform i.e frequency domain. In our work we have worked on matlab database images and natural images with noise intensities varying from 0.01 to 0.6, three types of noises i.e Gaussian noise, salt & pepper noise and speckle noise and different spatial techniques for enhancing the image. Thus, the contribution of this paper is to various techniques to enhance an image in spatial domain.
Top2. Noise Models
An image can be prone to numerous undesirable information called noise which can be added into it during any of the image processing methods to term a few it can be through acquiring of an image which may be due to device or memory failure, or surrounding atmosphere (i.e image capturing device are not proper, lenses being misaligned, focal length being weak, location of an image is not appropriate, scattering of storage, foggy weather conditions) or during transmission of an image. This Noisy image can be showed as in below equation (1) (Rafael et al.,2008):
g (x, y) = f (x, y) + n (x, y)
(1)where g (x, y) depicts for image with noise
f (x, y) depicts for given image
n (x, y) depicts for noise
2.1 Categories of Noise Model
Noise in an image represent the arbitrarily varying of pixel values, which is based on the probability of distribution. Probability Density Function (PDF) is used to indicate the noise model in an image. Identification of noise becomes very important pre-processing step in order to apply a suitable filter to eliminate the identified noise which helps in further processing of the image for the specified application like pattern recognition, object identification or any other application of interest. In our research we worked on three noise types namely impulsive noise called salt and pepper (S&P) noise, additive type noise namely Gaussian noise and multiplicative type noise like speckle noise. Hence, the noise model of these three have been presented along with their equation and Probability Density Function. (Yelmanov, S., & Romanyshyn, 2021; Al-Mahmood, H., & Al-Rubaye, Z., 2014).
2.2 Salt and Pepper Noise
The researchers Srinivasan et al. (2007), Church et al. (2008), Harikiran et al. (2010), Chen et al. (2010), Hanji et al. (2013), Koli et al. (2013), Hosseini et al. (2013) worked on impulse noise removal in an image. This type of noise appears during transmission of the data which has sharp and rapid changes in the image signal, this noise is randomized throughout the image with either bright pixel with a value of 255 (salt) or dark pixel with a value of 0 (pepper) or both the pixel values. The Probability density function equation is as in below equation (2) (Hiremath et al.,2021):
P(g) = P
a for g=a
P
b for g=b
(2)
0 otherwise
Probability density function of impulsive noise called Salt & pepper noise model (Hiremath et al.,2020), and the image are shown in figure 1