IMAGE QUALITY ANALYSIS INTRODUCTION

The term Image Quality Analysis is coined in this report to mean the investigation/extraction of certain useful information needed to describe image quality. For the purpose of a better understanding, some definitions are required and are thus given below; conversely image quality is further detailed. Imageis a representation of the external form of a person or thing in sculpture or screen display i.e. three-dimensional (3D), painting or photograph i.e. two-dimensional (2D) etc. Digital Imagethis could be referred to as a numeric representation of a twodimensional (2D) image. It may be of a vector or raster type depending on whether or not the image resolution is fixed. In this report the raster type is emphasized. Raster Imagesrefers to images which have a finite set of digital values, called pixels. A pixelalso known as picture element; is the single point in a raster image or the smallest unit in picture. Resolutionwhen referring to images, it is the number of pixels per inch i.e. dot per inch. This may depend on the camera settings.

IMAGE QUALITY This is a feature of an image that measures the perceived difference in an image in relation to an original/ideal image and it is the presence of distortion in an image that makes image quality assessment paramount. Image quality may depend on these listed points;

The deviation of the image formation of the camera from the ideal model, since the ideal model is only an estimate of the image information process. Image information is obtained typically as a result of light rays from the describe scene allowed to pass through the camera aperture to the image plane.

The difference from the original image when compressed stored or transmitted and decompressed again to retrieve the original image. The quality of the image measurement process based on the compression method used considering a large set of images. The coding artifacts introduced by JPEG coding method on the image produced by digital cameras.

IMAGE QUALITY ASSESSMENT Image Quality can also be perceived via 3 techniques

By objectively employing technical measures to determine the image deviation from its ideal model. The techniques used to objectively evaluate the quality of an image are; the Full Reference Image Quality Assessment (FR IQA) which utilizes an ideally certified reference image to evaluate the quality of the image under test, and the No Reference Image Quality Assessment (NR IQA) which requires no reference to access the quality of the test image. The subjective perception of image that is via the natural human eyes picturing a photograph. Image quality can be viewed from an angle to mean that the noise in the image may not reduce the quality if the noise does not affect the information which the viewer seeks to perceive. Verifying the contrast intensity of an image. A good image normally covers the entire dynamic range of the intensity levels.

IMAGE QUALITY FACTORS Certain factors influence the pre-processing and the post-processing of digital image and are listed below; Sharpness Noise Dynamic range

Colour accuracy Contrast Exposure accuracy Distortion Light fall offs Artifacts

OBJECTIVE In this report we shall aim to assess the image quality of a JPEG image taken from a Nokia E72-2 camera phone, of which three approximately same images were taken by setting the camera resolution to high, medium and low and also we shall assess the image quality of approximately same images taken by different phones to include BB 8520, Nokia C7-00, Nokia E55-1, Nokia E72-2 and Samsung GT-S8000.

JPEGJoint Photography Experts Group This is a widely used image file format for storing and transmitting digital images on computer systems and the internet respectively. It utilizes the lossy compression method to process image information i.e. some information are lost and may not be recovered by lossy information is known to give high compression ratio which guarantees a better image quality.

EFFECTS OF JPEG COMPRESSION The effect of JPEG is generally characterized by the artifacts it deposits on its image although it blends well into the image with detailed nonuniform textures making it difficult to see by human vision thus allowing higher compression ratios.

Since JPEG works by analyzing the patterns in regular blocks of the image and removing the 'less important' information. The main sign of compression is blocking and the blocks are most visible in areas of nearly flat colour and around the edges of things in the image where the colour changes significantly. Higher compression ratio affects the high frequency textures in the upper left corner of the images and very high compression ratio eventually affects the image quality of the image. The colour transformation used at the beginning of the compression process is justified since extracting the luminance component before down-sampling preserves the precision of the luminance.

PHONES CAMERA SPECIFICATIONS The specifications given are those made available by the manufacturer. Nokia E72-2

5 megapixel camera (2592 x 1944 pixels) with autofocus and flash Image formats: JPEG, EXIF CMOS sensor 5 x digital zoom One-touch autofocus key Focal length: 4.7 mm Focus range: 10 cm to infinity LED flash

BB 8520

2.0 megapixel camera (1600 x 1200 pixels) Image format: JPEG 5 x digital zoom

Nokia C7-00

8.0 megapixel camera (3264 x 2448 pixels) fixed focus Image format: JPEG Fixed focus Dual-LED flash Digital zoom

Nokia E55-1

3.2 megapixel camera (2048 x 1536 pixels) fixed focus Image format: JPEG, EXIF Fixed focus LED flash 4 x digital zoom

Samsung GT-S8000

5.0 megapixel camera (2592 x 1944 pixels) autofocus Image format: JPEG Autofocus Dual-LED flash Wide dynamic range

METHODS/TECHNIQUES The three techniques shall be used to determine the image quality of the JPEG image from the above listed cameras and conclusions shall be drawn from

each method while a general conclusion shall be obtained based on the three techniques.

1) CONTRAST INTENSITY The contrast intensity is verified by plotting the Histogram of the image which is design to take all the values in the grayscale image. First we shall consider the image taken by varying the resolution of the Nokia E72-2 to High, Medium and Low. We therefore read the images into MATLAB and plot the Histogram. The resulting figures are giving below;

x 10

0 0 50 100 150 200 250

Fi gure 1: High Resolution JPEG Image

Fi gure 2: Histogram of High Resolution JPEG Image

x 10

2.5

1.5 1

0.5

0 0 50 100 150 200 250

Fig Fi gure 3: Medium Resolution JPEG Image ure 4: Histogram of Medium Resolution JPEG Image

3000

2500

2000

1500

1000

500

0 0 50 100 150 200 250

Fi gure 5: Low Resolution JPEG Image

Fi gure 6: Histogram of Low Resolution JPEG Image

DISCUSSION In figure 4, 5 and 6 the histogram representing the contrast intensity of the High, Medium and Low JPEG image is viewed to be of good quality since the image covers the entire dynamic range of the intensity levels. However giving a close look at these histograms we could observe that the distribution of the contrast intensity looks the same but a reduction of the number of pixels has amounted to a reduction in the density of the Histogram plot as the camera resolution goes from High to Medium and eventually to Low.

Secondly we shall consider the image taken from the various the 5 phones listed above.

BB 8520

16000 14000 12000 10000 8000 6000 4000 2000 0 0 50 100 150 200 250

Figure 7: BB 8520 JPEG Image

Fi gure 8: Histogram of BB 8520 JPEG Image

Nokia C7-00

8 7 6 5 4 3 2 1 0 0 50 100 150 200 250

x 10

Fi gure 9: Nokia C7-00 JPEG Image

Fi gure 10: Histogram of Nokia C7-00 JPEG Image

Nokia E55-1

x 10 3

2.5

1.5

0.5

0 0 50 100 150 200 250

Fi gure 11: Nokia E55-1 JPEG Image

Fi gure 12: Histogram of Nokia E55-1 JPEG Image

Nokia E72-2

x 10

0 0 50 100 150 200 250

Fi gure 13: Nokia E72-2 JPEG Image

Fi gure 14: Histogram of Nokia E72-2 JPEG Image

Samsung GT-S8000

x 10 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0

50

100

150

200

250

Fi gure 15: Samsung GT-S8000 JPEG Image

Fi gure 16: Histogram of Samsung GTS8000 JPEG Image

DISCUSSION

BLIND IMAGE QUALITY ASSESSMENT - (BIQA) BIQA refers to the problem of evaluating the visual quality of an image without any reference, therefore this type of Image quality assessment falls under the NFIQA (No Reference Image Quality Assessment). REASONS OF BIQA BIQA is important because;

The use of FRIQA (Full Reference Image Quality Assessment) tends to address the fidelity of the image rather than the quality. In Image interpolation, the original high resolution image is often not available. In digital cameras, the original raw image is not accessible rather the compressed Q factored JPEG image is taken as original.

BLIND IMAGE QUALITY INDEX - (BIQI) This is a type of NFIQA that verify image quality by utilizing (NSS) natural scene statistics rather than the well-known pattern of other NFIQA which assumes that the distortion affecting the image is known. NSS is useful for defining the behaviour of an ideal observer in a natural task, typically be incorporating signal detection theory, information theory, or estimation theory.

In BIQI, the NSS is employed a 2 - stage image quality assessment emerges. First, distortion identification based on distorted image statistics (DIS) is undertaken in order to gauge the primary distortion affecting the image from a pool of possible distortions then secondly, distortion-specific quality assessment (DSQA) is then carried out to predict the quality of the image.