Possible Ways of Improving Imaging - Research Proposal Example

Image processing Your name Name of Assignment 25th July, 2014 Abstract Although edge detection has rapidly grown to become an invaluable tool in image processing, the technology still requires several improvements in order to enhance its reliability and efficiency. Sobel and canny have received much attention and intensive research on improving the image quality. Intensive research on filtering of an image has led to development of various image reconstruction techniques with an aim of reducing the noise while improving the image quality. This paper aims to study possible ways of improving imaging by filtering edges through the use of a filter, a selected filter, sobel and canny. From the analysis, it was found out that proper image reconstruction techniques like filtering, sobel, and canny improve imaging. Table of Contents Abstract 2 Table of figures 4 Introduction 5 Methodology 5 Results and discussion 6 Image detection 6 Converting image to from 3D to 2D 7 Selected filter 8 Rotating filter 45o 10 Rotating filter 90o 12 Rotating filter 135o 13 Comparison with Canny edge detector 14 Comparison with sobel edge detector 16 Conclusion 17 References 18 Appendix 19 Table of figures Figure 1: Image adopted from The Home Sitter 7 Figure 2: rgb2gray 8 Figure 3- filtered 10 Figure 4: filtered by filter45 11 Figure 5: filtered by filter90 13 Figure 6: filtered by filter135 14 Figure 7: filtered by canny 15 Figure 8: filtered by sobel 17 Introduction Imaging processing has been characterized by significant technological advancement thus leading to development of more sophisticated and efficient imaging modalities. These technologies provide detailed useful information of an image. The most common imaging processing is edge detection which is essential in data compression, image reconstruction and image segmentation as well as in matching. Among the image edge detection that has become the most favorable and outstanding are sobel, canny and filtering. Filtering is regarded as an invaluable in imaging technology because of its ability to produce images of outstanding quality non-invasively with no adverse side effects. There are several factors and conditions that must be met in order to obtain a high quality image. When these conditions and factors are not observed, edges are detected. Edges in images are simply pixels that do not represent the real state of image being examined. In such a case, the underlying image is visualized but other signals that do not relate to real image at that location are also present on the image. Although most edges have a tendency of degrading the image quality or rendering the image unusable, sometimes these edges may not be very visible and may be ignored. Thus, it is important to gain a proper understanding of edge detection, their characteristics and how they can be either be prevented or compensated. Methodology In filtering, the image is placed in the K-space such that the data associated with low frequencies is put at the center while those with high frequencies are placed around the center. Low frequency signal has more information about the signal (noise) and contrast. The k-space path followed in imaging sequence must cover all the data points in order to create a complete image. Sobel and canny imaging techniques are easier since their codes are simple to writer and they are inbuilt thus can not cause huge edge that can distort the image quality. When carrying out edge imaging it is important to accurately establish the Matlab code effects of in order to achieve reliable image reconstruction. The coding effects of the receiver sensitivities can be determined by either calibration before starting any examination or by obtaining the data during each image acquisition. When creating a 2-D image, significantly reduces the sampling time by minimizing the phase encoding steps. To achieve this, the distance between successive phase encoding lines is increased but the maximum k-space value is maintained. The reduction in number of K-space samples relative to the full sample is described using the reduction factor R. When filter imaging techniques are used, any reduction in the k-space sampling density results to a reduction in edge hence leading to detection. Results and discussion The image that was used is that of a car number plate were acquired to edge detection. Image detection >> I = imread('D:\Documents and Settings\Administrator\My Documents\house.JPG'); this will help in importing the image to work space figure Figure 1: Image adopted from The Home Sitter imshow(I); - displaying the images This plots the image that is the original as shown above; Converting image to from 3D to 2D The mage is made ready for processing using the following code BB = rgb2gray (I); will change the color of the image figure imshow(BB) - displaying the images The reconstructed image of without a filter, image displayed is shown in figure 2 Figure 2: rgb2gray Selected filter K-space filtering is one of the techniques that can be used to improve the quality of an image. The effect of various filters on the K-space can be studied using basic programming in MATLAB or any other programming application. When a low-pass filter is used on the K-space, there is no change in image contrast. However, the details and contours of the objects fade away. The disappearance of these details is attributed to the exclusion of the high spatial frequencies. On the other hand, the use of a high-pass filter results to an image with only the contours. This is because low spatial frequencies are eliminated. Low pass filter is commonly used for noise reduction while a high pass filter m may be used separate edges from image signal. The low-pass and high-pass filters can be combined to prude band-pass filter. The effect of the band-pass filter on the image can also be studied using MATLAB. The reconstructed image with filter and the is shown in figure 3. The following filter was selected Filter= Looking at the above filter, one can note that it has middle column unique. This type of filtering can also be obtained using g =imfilter(BB,[filter]/2,'replicate'. However, g =imfilter(BB,[filter]/2,'replicate' does require coding. The image for the house shown in figure 3 below was used. >> filter=[1 1 0 -2 -2;1 1 0 -2 -2;1 1 6 -2 -2;1 1 0 -2 -2;1 1 0 -2 -2]; >> g =imfilter(BB,[filter]/2,'replicate'); this filters the image at 0o mask figure imshow(g) Figure 3- filtered In figure 3, the loss of the some edges is attributed to the non-uniform-valued filter. This means that the superposition for every image is achieved by adding different weighted pixel values. The edges are located in an area with a high SNR so that the effect of observation noise on the fit is significantly reduced. In order to achieve an improved reconstruction, filter has used several neighbouring acquired k-space data points to interpolate missing points. Rotating filter 45o When the filter is rotated at 45o and entered in to Matlab it coded as filter45=[0 -2 -2 -2 -2;1 0 -2 -2 -2;1 1 6 -2 -2;1 1 1 0 -2;1 1 1 1 0]; g45 =imfilter(BB,[filter45]/2,'replicate'); this filters the image at 45o mask figure imshow(g45) The rotated filter produces reconstructed image of as shown in figure 4 below Figure 4: filtered by filter45 Images reconstructed using filter rotated 45o appears to have more noise than image generated using the filtered, probably because filter trades noise reduction for the removal of other edges. The figure above shows that the brightness or darkness is equal that is 50% this means that a rotated filter at 45o separates right feature of an object or image by showing grayscale intensity. This ensures that the image background values and contract adjustments are well pronounced. The effect of rotated filter at 45o is illustrated in the e ray above. Dark features of a bone are shown clearly while right features are pronounced. It helps to obtain a thin edge of grey level image instead of binary image. It maximizes edge information fields by having a minimum allowance for excess or missed edges. In the image above the information of the original image has been changed by minimizing edge information to produce a filtered image. Rotating filter 90o Rotated imaging reduces the number of phase-encoding steps but still generates images of acceptable quality. >> filter90=rot90(filter); totating the filter 90o >> g90 =imfilter(BB,[filter90]/2,'replicate'); filtering the image >> figure imshow(g90) plotting and displaying the image Figure 5 shows our reconstructed image using a filtered rotated 90o with relation to reference image. Figure 5: filtered by filter90 It is worth to mention that the filter applied didn’t produce a clear image at the start, as it is seen in the figure due to aliasing noise. Filter rotated 90o produces a better image quality in comparison to the conventional filter, in order to prove that I chose to compare figure 3 and 5 that filter rotated 90o achieves a superior image quality than the filter. Therefore effectively removes noise experienced in the filter. It also preserves the resolution of the axial image without blurring. Slightly blurred images are obtained when using the angle, probably due to the trade-off between noise suppression and preservation of resolution. This technique significantly minimizing the noise and the echo time by moving the echo away from the center in the acquisition window. Rotating filter 135o >> filter135=rot90(filter45); >> g135 =imfilter(BB,[filter135]/2,'replicate'); >> figure >> imshow(g135) Figure 6: filtered by filter135 The image manipulated by filter rotated 135o shows effect of an image that the energy that is contained in the image edges will end up leaking from the signal frequency bin into the other frequency bins that are adjacent to it. The result of this leakage is errors in the amplitudes of the displayed. Due to these amplitude errors, the overall effect will be a case whereby pixels that occur very close to larger ones will end up being obscured. Thus, in order to reduce the effects of this edge, a process known as windowing which basically utilizes window functions can be used. The image manipulated is shown figure 6. Comparison with Canny edge detector The image was read into work space as follows can = edge(BB,'canny'); specifies the Canny method >> figure imshow(can) Figure 7: filtered by canny The reconstructed image using canny function had their color change as well the edges disappeared. The image had a lot of light meaning that the filtered allowed light to go through. The image was more blurred than in other cases as shown above. This means the filter reduced the sharpness of the image and the colors of the image. This filter is used in modeling optical length which can be used in blurring objects. It is also used in designing equipments which can reduce noise of the image. If the image obtained after filtering is added to the original image a clear and sharper image will be made. Comparison with sobel edge detector The sobel image detector usually finds its application in areas where short acquisition times are required. Generally, they reduces noise because shorter readout trains minimize various unwanted effects that can distort the image quality and can also be used to enhance spatial resolution. >> sobe = edge(BB,'sobel'); shows the method used >> figure >> imshow(sobe) Figure 8: filtered by sobel When sobel operation was used on the image, the pixels were detected. The images produced and viewed using binary blocks where the grey level of the image has been turned into black. The blackness that is pronounced is as a result of binary block configurations within the edges. When a sobel function is applied to an image, filtering takes place compromising the image detail and computational complexity which turns blocks to be large thus increasing the configurations. Small configurations are minimized becoming insignificant. The filtering of the image using sobel functions helps eliminate grey parts of the image. It is important at this point to note that sobel is critical in film production as some images will be produced after eliminating edge information. Conclusion In this paper the working principles of image processing were discussed in details. This discussion included the rotating of a filter 45, 90 and 135 degrees, canny and sobel techniques in edge detection. It was found out that by using MATLAB programming parameters of edge detection can be varied in order to obtain the desired image quality. References Bergenudd, E., 2006, Low Cost Real Time License Plate Recognition for a Vehicle PC’, Master’s Degree Project. Gonzalez RC & Woods, RE, 2008, ‘Digital Image Processing, Instructor's Manual’ New York: Prentice Hall. Gonzalez, RC, Woods, RE and Eddins, SL, 2009, Digital Image Processing Using MatLab’, New York: Prentice Hall. The Home Sitter,, 2014. Best House Designers- Great house designers – 2250 sq ft 4 bhk double storey house design Appendix Read More