Lecture - Digital Image Processing

Digital Image Processing

Overview
1. Why use Digital Image Processing
2. Digital Image Fundamentals
3. What is Digital Image Processing


Why use Digital Image Processing

Digital image processing provides a new flexible environment that can be used to experiment to achieve a desired effect. With digital image processing there is now a much greater variety of image manipulation, transformation and enhancement options that are available. These new options were never as freely accessible with the old darkroom photography developing. Since digital processing has become more widely known this means the number of effects is always increasing.

An image capture system contains a lens and a detector. This detector is often a CCD and is in a linear array or a matrix array of photosensitive electronic elements. 

 

source - http://photocafe.pro/wp-content/uploads/2006/06/ccd.jpg

On an area array sensor there is a grid that is made up of hundreds of thousands of tiny microscopic cells. Each individual cell is less than 5μm in size. Each cell creates pixels by sensing the intensity of the light formed by a lens system. When the resolution of a picture is too low then the picture will appear pixelated and blocky. Most modern devices will have a sensor with more than 1 million cells.

A digital camera stores and captures images in 3 different colours ; red, green and blue. Each cell is assigned a three 8-bit number meaning that each cell has 256 levels. The light is collected by the lens and then it is focused onto the cameras sensor array. While this is happening the light is passed through several filters that remove data that is beyond the resolution of the sensor, compensate for false or incorrect colourations that are caused by colour contrasts and finally reduced the infrared and non visible light levels that disturb the imaging process.

When taking a picture sometimes a pattern of lines or waves will appear on objects or materials in a picture. This occurs when a fiber on an object match the chip inside the camera. However companies incorporated anti-aliasing filters into the cameras so that this would blur the small details but other companies prefer to not use these as it sacrifices the sharpness and crispness of the pictures that the camera take.

Digital Imaging Fundamentals

Pixels 

Digital images are called raster-scan or bitmap images. Each image is made up of a grid of pixels. Each individual pixel is in a path of colour but the display itself is made up of red, green and blue phosphor dots or stripes.

source - http://www.dansdata.com/images/io026/pointer_close.jpg

A pixel is the smallest digital element that is available when using image editing software, Each pixel can be coloured but because the pixels are only an approximation of the colour of the subject then the bit mapped images can sometimes become blocky and have jagged edges. This is the downfall of bit mapped images but they are small in memory size so are some times the better choice. A bit mapped image is represented as an array of groups of bits. Each group codes the colour of the single pixels on the screen. The array could be at 640x480 resolution with each pixel having 24 bits in each group.

640 x 480 x 27 = 7,372,800 bits
7,372,800 / 1024 / 1024 = 7.4Mb

Dynamic Range

The dynamic range is the number of colours or shades of grey. The dynamic range of a digital image, however, is fixed by the bit depth of each pixel. This determines the maximum number of shades of colours of grey that can be accessed in the software's palette.

source - http://bytechunk.net/quake2/quake2_palette.jpg

Bit Depth

A 1 bit pixel will only be able to display 2 values. Black or white. To represent grey pixels in an image a process called half-tone is used. This gives the impression of shading where there is none.

source - http://sweetclipart.com/multisite/sweetclipart/files/halftone_pattern_black_white_1.png

An 8 bit pixel will be able to display up to 256 shades of grey (2^8 = 256) With this number of values it means that representing shading in an image. This shows a more gradual level of shading which is more realistic than only 2 values. 

source - http://upload.wikimedia.org/wikipedia/commons/9/9a/Colour_banding_example01.png

A 24 bit pixel will be able to display > 16 million colours (2^24 = 16million). This is called true colour and each pixel will have varying levels of red, green or blue. This will mean that unlike the previous pictures, where shading was implied or was more blocky and a bit more obvious, a 24 bit picture will be crisp and clear. 

source - http://file.answcdn.com/answ-cld/image/upload/h_320,c_fill,g_face,q_60,f_jpg/v1400839612/s2yb1qiscgsq4gjvucoj.png

Colour Palette

The computer will predetermine the colour palette that is used by the system ie 256 colours will be used for every image. The fidelity of an 8 bit image is enhanced by the selection of the 256 colours that will be used for the image. This is known as adaptive palette but the problem is that it can cause problems when multiple images are being displayed at once on a system that can only display 256 colours at one time. 

source - the lecture slides

source - the lecture slides

Below is an example of an optimised and unoptimised palette.

source - the lecture slides

What is Digital Image Processing?

Digital Image Processing can be separated into 4 main categories. 

Analysis - Operations will provide information about photometric features such as a histogram and the colour count.

Manipulation - This operation is about changing aspects of the picture using tools such as fill and crop.

Enhancement - Functions that will  be for enhancement will be to improve the quality of the image with operations such as edge enhancement and heighten contrast.

Transformation -  As the name suggests the operation is for changing the image with rotation and skew operations.

Analysis

A common operation is the histogram display which plots the various intensity levels such as 8 bit grey scale levels ranging from 0 - 255 on the horizontal axis and then the number of pixels having each level on the vertical axis. 

source - the lecture slides

Transformation

The rotation of image is part of the transformation operation. The pixels are simply remapped.

source - the lecture slides

Manipulation 

The area that is to be changed will be highlighted or selected using a selecting tool such as the lasso and once this area is selected it can then be change and modified. In the case below the selected area is filled in with an orange colour.

source - the lecture slides

Enhancement

An example of enhancement is filtering. This is when the kernel takes the pixel value and multiples it and then totals the value giving the pixel a new output.

source - the lecture slides

Another example of enhancement is blurring which is when the surroundings of an image can be simulated to look blurred to highlight an important part of a picture of a key point.

source - the lecture slides