Approaches For The Image Denoising

Images are electronic snapshots taken of a scene or scanned from documents, such as photographs, manuscripts, printed texts, and artwork. The digital image is sampled and mapped as a grid of dots or picture elements (pixels). Each pixel is assigned a value resulting from the illumination intensity which is represented in binary code (zeros and ones). The binary digits ("bits") for each pixel are stored in a sequence by a computer.

Images are used in diverse applications such as remote sensing, medical image analysis, content-based image retrieval and visual quality inspection has led to an increased interest in image processing. To get an enhanced image or to extract some useful information from an image one needs to process the image. Image noise is a random variation in image brightness or is unwanted signal that interferes with the original signal and degrades the visual quality of digital image. The main sources of noise in digital images are imperfect devices, problem with data acquisition process, interference by natural phenomena, transmission and compression which reduces image quality.

The image noise could be either additive or multiplicative. For an image to be analyzed through a stage of restoring its details known as the image processing. Image processing are actions performed on the images to overcome problems with them in preparation to introduce them to an algorithm for a specific task or viewing. The aim of image processing is to enhance the image data by suppressing unwanted distortions and enhancing some image features using image restoration, image enhancement and image denoising techniques. Image processing improves of pictorial information for human interpretation, Compression of image data for storage and transmission, Preprocessing to enable object detection, classification, tracking and noise Reduction.

Image denoising is the process of reconstructing a signal form noise therefore removing unwanted noise in order to restore the image. The main goal of image denoising is to remove noise and preserve useful information. There are different techniques in filtering out noise from an image which are categorized into spatial and frequency domain.

The spatial domain filtering deals with the image matrix in which the processed value of the current pixel processed value for the current pixel depends on both itself and surrounding pixels. While the frequency or transform domain the image is further transformed, multiplied with the filter function and then re-transformed into the spatial domain.

Images are important to humans because a large part of the brain is focused on visual processing, allowing humans to process images at an incredible speed. The human brain can recognize a familiar object in only 100 milliseconds, making images play an important role in communication in today's short-attention world. Typical application areas of images include Television Signal Processing, Satellite Image Processing, Medical Image Processing, Robotics, Visual Communications, and Law Enforcement.

Naturally images contain some amount of noise. Noise which is also known as an unwanted electrical or electromagnetic energy that degrades the quality of signals and data. Image noise is one the most common degrading factors which affects the visibility of images and makes them unclear. Such degradations are as a result of image acquisitions due to low grade image sensors, electronic interference or during transmission due to electronic interference in transmission channel.

Generally, noise can be categorized into additive where noise signal gets added to the original signal to produce a corrupted noisy image and multiplicative where noise is multiplied into the original signal to produce a noisy image.

Different techniques have been proposed till date in order to filter out noise from images. Filtering is a technique for modifying or enhancing an image, to filter an image means to emphasize or remove other features. Noise Filtering techniques employ the use of linear and non-linear methods in the spatial domain, the linear method uses the mean while the non –linear involves the use of order statistics. The non-linear method which we would be using in this thesis also exploits the use of median filter is the earliest and most popular that uses aggregated ordering, the adaptive median The filter kernel coefficients change their values according to the image structure, which is to be smoothed, the weighted A non-negative weight is assigned to every pixel inside the filtering window offering more flexibility and also the switching also known as decision based median filter checks each input pixel whether it has been corrupted by noise or not.

Various approaches for the image denoising detects and filter a single noise type on an image. Since images can contain noise from acquisition stage and also be transmitted, during the transmission another type of noise can be added onto the image. Removal of a specific type of additive noise ( salt-and-pepper noise, Gaussian noise, impulse noise, shot noise, random noise, quantization noise) or Multiplicative noise (speckle noise). Most filters are designed for a specific noise type, there is need to have an approach to detect as many as possible different noise types. During testing the authors use different images for different denoising algorithms.