Sei sulla pagina 1di 4

PROPOSAL

I
TITLE: IMAGE SEGMENTATION TECHNIQUES USING MATLAB: A COMPARATIVE STUDY

I.

INTRODUCTION and MOTIVATION Image segmentation deals with the concept of how to isolate objects or parts of objects from the rest of the image. For example in many surveillance images/video frames, we might be interested in getting the information of particular object e.g. people, cars etc. Rest of the regions e.g. background, might not be of our interest. Image segmentation techniques come handy, to achieve the goal of getting the region of interest (ROI). Also there are many situations where we want to separate out parts of objects, such as isolating just the face or the hands of a person. We might also want to preprocess an image into meaningful super pixels, which are segments of an image that contain things like limbs, hair, face, torso, tree leaves, lake, path, lawn and so on. Using super pixels saves on computation; for example, when running an object classifier over the image, we only need search a box around each super pixel. We might only track the motion of these larger patches and not every point inside. Beyond, this image segmentation also helps in image reconstruction (by filling the holes) and in-painting as well. Hence, the process of getting an accurate segmented image has been an exciting field of research for many decades.

II.

SCOPE OF THE PROJECT The project is intended to introduce various steps involved in image segmentation by using different methods, available in literature. The project will start with the discussion of how a simple method can be used to get a segmented image by using gradient and laplacian edge detectors. Since, they might not result in well-defined gross representation of the ROI, some more sophisticated techniques involving basic image operations and processing using morphology will be discussed. This will involve: erosion, dilation, opening, closing, thresholding and their practical applications in noise removal, character / License Plate Number segmentation etc. Further, some classic morphology algorithms e.g. region filling, connected components and skeletisation will be discussed. Further the project will

Page 1 of 4

lead to discuss other complex algorithms for image segmentation e.g. Watershed algorithm, image repair by in-painting and mean-shift segmentation.

III.

NOVELTY OF THE PROJECT The novelty of the project will lie in proposing a new image segmentation algorithm, combining the effects of any of the classic methods (stated above), with the RGB or HSV color model. The results will include the comparison between the performance analyses of the respective existing image segmentation models with the new one.

IV.

METHODOLOGY i) ii) The project will be implemented by studying related research papers. Test image segmentation data set will be used from: Berkeley Segmentation Dataset: Images available at http://www.eecs.berkeley.edu/Research/Projects/CS/vision/bsds/BSDS300/ht ml/dataset/images.html. Implementation details: System Requirements: Intel (R) core (TM), i5-3210M CPU@2.50 GHz. 6GB (RAM), 64 bit Operating System. Software Requirements: Microsoft visual studio 2010, OpenCV2.1 and MATLAB.

iii)

V. i) ii)

REFERENCES Gonzalez, Rafael C., and E. Richard. "Woods, digital image processing." ed: 3, Prentice Hall Press, ISBN 0-201-18075-8 (2002). Zhang, Yu J. "A survey on evaluation methods for image segmentation." Pattern recognition 29.8 (1996): 1335-1346.

Page 2 of 4

PROPOSAL II
TITLE: DIGITAL WATER MARKING - SECURE ENERGY EFFICIENT POWER SPECTRUM ANALYSIS

I.

INTRODUCTION and MOTIVATION

Usage of digital media has witnessed a tremendous growth from desktops to laptops and now to the hand-held devices, as a result of their notable benefits in ease of manipulation and transactions from anywhere, at any time. Future will also include its colossal usage in the field of banking, data biding, auctions, military data transmission, etc. This in turn demands high level of safeguard and digital authentication as technologies, such as the internet, broadband availability and mobile access advances. However such digital media is progressively vulnerable to copyright infringement, tampering and unauthorized distribution, there by declining huge business profits. The protection of digital information has received significant attention within the engineering community. Digital watermarking technique, which identifies the source of content, has been proposed as a possible solution to this problem.

II.

SCOPE OF THE PROJECT

An efficient watermarking system can be developed by testing the features of proper domains: DCT or DWT, in which watermarks can be inserted or retrieved. However, one of the solutions to find an optimal way of embedding the hidden information in an image to overcome the distortion premised by various attacks may involve usage of the multiresolution wavelet decomposition. Wavelets are functions that satisfy certain mathematical requirements (e.g. localized in time and frequency, integrate to zero, quick and easy to calculate the direct and inverse wavelet transforms) and are used in representing data or other functions. Wavelet algorithms process data on different scales and resolutions. If we look at a signal with a large window, we would notice gross features. Similarly, if we look at the signal with small window, we would notice small features. The result in wavelet analysis is to see both the forest and the trees. III. NOVELTY OF THE PROJECT

This project work, will concern with resisting attacks, such as Geometrical Transformation, Compression, Additive Noise, etc on digital watermarking. A contrarian approach of placing the watermark in the perceptually significant component of the image will be dealt with. An important open problem is the construction of a method that would identify perceptually significant components from an analysis of the image and the human perceptual system. Thus, it
Page 3 of 4

will be proved that the choice of the DCT is not critical to the algorithm and other spectral transforms, including wavelet type decompositions, can also produce good results. No noticeable distortion can be seen in the watermarked images. However, a significant attention will be paid to which wavelet may be preferred. By studying the daubechies (db) wavelets at various levels with different levels of decomposition can also be analyzed. And, finally the most optimal solution for insertion of the watermark can be determined. IV. i) ii) iii) METHODOLOGY The project will be implemented by studying related research papers. Test image data set will be used from: tid2008 and databaserelease2 Implementation details: System Requirements: Intel (R) core (TM), i5-3210M CPU@2.50 GHz. 6GB (RAM), 64 bit Operating System. Software Requirements: MATLAB. REFERENCES Gonzalez, Rafael C., and E. Richard. "Woods, digital image processing." ed: 3, Prentice Hall Press, ISBN 0-201-18075-8 (2002). Raghuveer M. Rao , Ajit S. Bopardikar. Wavelet Transforms: Introduction to Theory & Applications. ed: Har/Dskt, Prentice Hall Press, ISBN 0201634635 (1998).

V. i) ii)

Page 4 of 4

Potrebbero piacerti anche