Slaven Marusic, David B.H.Tay et al presented a elaborate survey of biorthogonal ripples in digital watermarking. In this paper they derived biorthogonal ripple coefficients utilizing Cohen-Daubechies- Feauveau ( CDF ) biorthogonal ripple system.

## Nagaraj.V.darwadkar Method

Nagaraj V. Dharwadkar et Al [ 6 ] , proposed a non-blind watermarking strategy for colour images in RGB infinite utilizing DWT-SVD in 2010. In this method, the water line is embedded into cover image in RGB infinite. The combinations of distinct ripple transform and remarkable value decomposition of bluish channel is used to implant water line. The remarkable values of different bombers band coefficients of bluish channel are modified utilizing different scaling factors to implant the remarkable values of the water line. The transcript of the water line is embedded into four bomber set coefficients which are really hard to take or destruct [ 6 ] .

## Watermark implanting process

Measure 1: Read the colour image I of size NxN.

Measure 2: Read the monochromatic image Ten of size MxM and use DWT on X to acquire D= { dij } of size MxM.

Measure 3: Compute R, G and B channels from colour image I of size NxN.

Measure 4: Transform R, G and B channels into Y, I and Q channels of the colour image.

Measure 5: Compute 3rd degree DWT on Y channel to acquire the frequence constituents { HH1, HL1, LH1, { HH2, HL2, LH2, { HH3, HL3, LH3 } } } .

Measure 6: Embed the water line frequence coefficients, get downing from HH1 for each row select the frequence coefficients in falling order with regard their absolute values.

Measure 7: Modify each frequence coefficient degree Fahrenheit of screen image to ij. If the subcomponent HH1 is deficient to implant the complete water line, so insert in the other coefficients in the order { HL1, LH1, { HH2, HL2, LH2, { HH3, HL3, LH3 } } } .

Measure 8: Salvage the location of the modified frequence coefficients into a cardinal array K of size NxN. The cardinal array consists of value 1 if the coefficient is modified otherwise 0.

Measure 9: Replace by in decomposed Y channel and compute opposite DWT of modified Y channel.

Measure 10: combine modified Y channel with I and Q to acquire watermarked image.

## Watermark extraction process

Measure 1: Read the watermarked image of size of size NxN.

Measure 2: compute, ‘ and channels of the watermarked image.

Measure 3: Transform these, and channels into, and channels.

Measure 4: Compute 3rd degree DWT on channel to acquire the frequence constituents { HH1, HL1, LH1, { { HH2, HL2, LH2, { { { HH3, HL3, LH3 } } } } .

Measure 5: Compute 3rd degree DWT on Y channel of the un-watermarked image to acquire the frequence constituents { HH1, HL1, LH1, { HH2, HL2, LH2, { HH3, HL3, LH3 } } } .

Measure 6: Extract the water line spots from the frequence subcomponents utilizing the cardinal array K as ij= ( – ) /I± . If ijE? T, so ij =1 other wise ij = 0.Whwere i= 1,2,3aˆ¦M and j= 1,2,3aˆ¦M.

## 2.3.2.2 Yanhong Zhang Method

Yanhong Zhang [ 7 ] proposed a unsighted water line embedding/extracting algorithm utilizing RBF nervous web. In this method the DWT is used to get the better of the blocking phenomenon jobs in DCT. First, the original image is 4-scale degree DWT transformed, and decided the watermarking strength harmonizing to HVS. When implanting water line, a secret key is used to find the water line get downing location, and after that, embed and pull out the water line by utilizing the trained RBF [ 7 ] .

## Watermark implanting process

Measure 1: Transform the original image utilizing DWT as is the LH4, HL4, HH4 sub-band coefficient.

Measure 2: Choose the get downing place of water line implanting coefficient utilizing the secret key.

Measure 3: Quantize the coefficient of DWT, ( i+key ) by Q, as the input to the RBFN and so acquire the end product.

Measure 4: Embed the water line harmonizing to the undermentioned equation ; Where is the water line sequence, Q is quantization value and is the coefficient of the watermarked image.

Measure 5: Perform IDWT to acquire the watermarked image.

## Watermark extraction process

Measure 1: Transform the watermarked image by the DWT transform as with the bomber set coefficients LH4, HL4, HH4.

Measure 2: Quantize the DWT coefficient by Q as the input to the RBFN and so acquire the end product.

Measure 3: Extract the water line utilizing the undermentioned expression.

Measure 4: Measure the similarity of the extracted water line and the original water line utilizing the equation

Measure 5: Use, threshold as a key to judge if there is an embedded water line or non. If is larger than threshold and the location is equal to identify, the water line is affirmed.

## 2.3.2.3. He Xu, Chang Shujuan Method

He Xu, Chang Shujuan [ 10 ] , proposed an adaptative image watermarking algorithm based on nervous web. In this method, the watermarking signal is embedded in high frequence, which is in the lower frequence of original image by DWT joined with DCT. The ability of pulling is improved by pretreatment and retreatment of image scrambling and Hopfield web [ 10 ] .

## Watermark implanting process

Measure 1: The watermarking signal is applied as the preparation signal input to the Hopfield web in order to complete the storage of the water line.

Measure 2: After making scrambling transform, the water line signal R is generated. The affine transform is used as scrambling transform, the key is scrambling times, and so the water line pretreatment is completed.

Measure 3: The low frequence sub- image LL is extracted from the original image by utilizing the first order DWT transform. I will be gotten by DCT transform which procedure 8×8 block breakdown.

Measure 4: The scrambling water line sequence is embedded in high-frequency coefficients of the image I harmonizing to the equationin order to acquire. Where is implanting strength in the scope 0 1.

Measure 5: The IDCT is performed to acquire the low frequence sub-image LL which contains water line and IDWT is performed to acquire the water line image.

## Watermark extraction process

Measure 1: The detected image and original image are processed by first order DWT and T and I are gotten through DCT barricading phenomenon.

Measure 2: Water line is extracted through T and I input watermark sensing faculty.

Measure 3: The extracted water line signal R is processed harmonizing to cardinal reverse scrambling to acquire the water line.

Measure 4: The extracted water line is applied as input to the Hopfield web and after informations treating the water line is extracted.

## 2.3.2.4. Charu Agarwal et Al Method

Charu Agarwal et Al, [ 34 ] proposed digital image watermarking in DCT sphere utilizing fuzzed illation system. In this method, Human Visual System ( HVS ) features are modeled utilizing a Fuzzy Inference System ( FIS ) for robust image watermarking. The fuzzed input variables matching to luminance sensitiveness, edge sensitiveness computed utilizing threshold and contrast sensitiveness computed utilizing discrepancy are fed to a FIS driven by 10 fuzzed illation regulations. The FIS produces a individual end product burdening factor which is used to implant a indiscriminately generated normalized water line with in the host image in the DCT sphere. The signed image has good perceptual quality and is capable to stir grade image processing onslaughts. The high computed value of PSNR indicates hardiness of the implanting algorithm. The water line is extracted from the signed image utilizing celebrated Cox ‘s algorithm [ 34 ] .

## Watermark implanting process

Measure 1: Cover image is divided into 8×8 blocks in spacial sphere DCT is computed on all blocks.

Measure 2: Compute border sensitiveness ( threshold ) , luminance sensitiveness and contrast sensitiveness ( discrepancy ) of all blocks of screen image.

Measure 3: Supply these threshold, discrepancy parametric quantities as input to fuzzy illation system.

Measure 4: Apply fuzzed illation regulations to the fuzzed illation system and obtain the water line burdening factor.

Measure 5: Perform water line implanting in low frequence DCT coefficients of screen image.

Measure 6: Calculate the IDCT to obtain the watermarked image.

## Watermark extraction process

Measure 1: Compute DCT of all 8×8 blocks of screen and watermarked ( signed ) images.

Measure 2: Subtract the computed coefficients of original image from watermarked image.

Measure 3: Recover the water line utilizing fuzzed illation system.

Measure 4: Compare the cured water line with the original water line utilizing Sim ( X, X* ) parametric quantity.

## 2.3.2.5. Sameh Oueslati et Al Fuzzy Method

Samesh Oueslati et Al [ 35 ] , proposed a fuzzy watermarking system utilizing the ripple technique for medical images. In this method, an adaptative watermarking algorithm performed in the ripple sphere is proposed which exploits a human ocular system ( HVS ) and a fuzzed illation system ( FIS ) . HVS is adopted to further guarantee the water line invisibleness. The FIS is utilized to calculate the optimal water line burdening map that would enable the embedding of the maximal energy and unperceivable water line. For the intent of security and hardiness, a water line sequence is embedded by selectively modifying the middle- frequence parts of the image [ 35 ] .

## Watermark implanting process

Measure 1: Input signal the screen image and watermark image.

Measure 2: Convert the water line into a watercourse of binary informations dwelling of nothing and 1s.

Measure 3: Decompose the host image utilizing Haar ripple transform.

Measure 4: Insert the information into ripple coefficients, which have the largest values in in-between frequence coefficients.

Measure 5: Perform the opposite Haar ripple transform to acquire the watermarked image.

Measure 6: Expose the watermarked image.

## Watermark extraction process

Measure 1: Input signal the watermarked image.

Measure 2: Decompose the watermarked image utilizing Haar ripple transform.

Measure 3: Choose the ripple coefficients which have largest values in in-between frequence bomber set.

Measure 4: Compare the coefficients of screen image and watermarked image depending upon the location.

If the coefficient of embeddingE? original coefficient so the informations shop in it is 1

If the coefficient of embeddingE‚ = original coefficient so the informations shop in it is 0

Measure 5: Expose the cured image.

## 2.3.2.6. Ming-Shing Hsieh Method

Ming- Shing Hsieh [ 36 ] proposed perceptual right of first publication protection utilizing multi-resolution wavelet- based watermarking and fuzzed logic. In this method, an expeditiously DWT-based watermarking technique is proposed to implant signatures in images to certify the proprietor designation and deter the unauthorised copying. This technique is based on using a context theoretical account and fuzzed illation filter by implanting the water lines in the larger entropy coefficients of coarser DWT bomber bands [ 36 ] .

## Watermark implanting process

Measure 1: Kind the Grey degrees of water line of size ‘n ‘ in go uping order to bring forth the sorted water line.

Measure 2: Decompose the host image into three degrees with 10 subbands of ripple pyramid construction and take a subband ( HL3 ) to implant water line.

Measure 3: Calculate the leaden information of coefficients.

Measure 4: Let the preset interval be and allow T be the figure of referenced coefficients used as a key to pull out water line without the host image. Coefficients with larger information are chosen from subband Where. The larger entropy coefficients make the water line more robust and transparent. If so otherwise Where is used to acquire integer portion of its statement. Let { } be the set of referenced coefficients and the coefficients to be embedded water lines ; { } is called the alternate coefficients. Screening { } to bring forth { } called the sorted alternate coefficients.

Measure 5: Quantize { } utilizing a preset interval, which will pull out the water line W without the screen image.

Measure 6: Embed water line SW into subband HL3 utilizing the equation

, To+T1+T2 ) /3=EnixT1.

Measure 7: Salvage the symbol of embedded subband and execute IDWT to acquire the watermarked image.

## Watermark extraction process

Measure 1: Decompose watermarked image into three degrees with 10 subbands utilizing DWT.

Measure 2: Restore the grading factor vi the symbol of embedded subband, symbol map of SCi, corresponsive map of Ci and SCi and corresponsive map of Wi and SWi.

Measure 3: Extract the sorted water lines by the proposed extracting watermarking algorithm.

Measure 4: Rearrange the water lines from corresponsive map of Wi and SWi to acquire the extracted water line.

## 2.3.2.7. Soheila et Al Method

Soheila Kiani et Al [ 37 ] , proposed Fractal based digital image watermarking utilizing fuzzed C-mean bunch. In this method a new watermarking method is used to implant a binary water line in to an image. The proposed method uses a particular type of fractal cryptography that its parametric quantities are contrast scaling the mean of fury block. Besides, it utilizes the fuzzed C-mean bunch to turn to the water line bits [ 37 ] .

## Watermark implanting process

Measure 1: The fractal encryption is applied on the original image to bring forth fractal codifications for all scope blocks.

Measure 2: Use the fuzzed C-mean bunch on all the blocks and sort them into four groups.

Measure 3: As per the centres calculated in old measure determine category A and B.

Measure 4: For each spot of water line:

If the spot is zero, P scope blocks that their matched sphere blocks are classified as A and selected indiscriminately.

If the spot is one, P scope blocks that their matched sphere blocks are classified as B and selected indiscriminately.

Measure 5: Fractal decryption procedure is used to build watermarked images.

## Watermark extraction process

Measure 1: Fractal cryptography is performed on watermarked images to bring forth fractal codifications of all scope blocks.

Measure 2: The fuzzed C-mean bunch is applied on all blocks to sort them into four categories.

Measure 3: Harmonizing to the bunchs category A and category B are determined.

Measure 4: For all scope blocks the water line spots are determined harmonizing to the secret key as follows

If the most matched sphere blocks belong to category A so the spot is zero.

If the most matched sphere blocks belong to category B so the spot is one.

If the most matched sphere blocks belong to category C or D so the spot is undetermined.

Measure 5: Perform measure 4 on all spots of watermarked image harmonizing to the secrete key.

These characteristics have motivated to develop two new methods for watermarking in transform sphere utilizing Back Propagation Neural Network ( BPNN ) and Dynamic Fuzzy Inference System ( DFIS ) .

## Research OBJECTIVES

The aims of this research work are as follows:

To research digital image watermarking techniques utilizing Back Propagation Neural Network and Dynamic Fuzzy Inference System in Discrete Wavelet Transform sphere.

To develop watermarking techniques, which are unperceivable for unauthorised user, without impacting the original image quality.

To develop blind watermarking techniques so that the water line can be detected without the original image.

To develop water line techniques, which are robust against cropping, salt & A ; pepper noise, rotary motion, JPEG compaction etc. , and holding domination over bing watermarking methods.

## PROBLEM STATEMENT

From the literature reappraisal, it is evident that the digital image watermarking can be achieved by utilizing either implanting the water line straight into the image pels of the screen image or into the transformed coefficients of the screen image. There are several demands that the implanting method has yet to fulfill. Making the robust and blind digital image watermarking methods is still a ambitious undertaking for research workers. These algorithms are robust against some onslaughts but non against most of them. Besides, some of the current methods are designed to accommodate merely specific application, which limits their broad spread usage. Furthermore, there are drawbacks in the bing algorithms associated with the watermark-embedding sphere. These drawbacks vary from system to system. Watermarking strategies that modify the LSB of the informations utilizing a fixed magnitude PN sequence are extremely sensitive to signal processing operations and geometric uses. This will restrict their usage in big figure of applications.

To enable copyright protection and hallmark, robust digital water line can be embedded into multimedia contents unnoticeably. However, geometric deformations pose a important menace to robust image watermarking because it can desynchronise the water line information while continuing the ocular quality. To get the better of this, the robust digital image watermarking strategy utilizing Back Propagation Neural Network in DWT sphere is proposed, in which the geometrical effects such as cropping and rotary motion are minimized. Back Propagation Neural Network has good nonlinear estimate ability. It can set up the relationship between original ripple coefficients and watermarked ripple coefficients by seting the web weights and prejudice before and after implanting water line. Owning to the usage of nervous web, we can pull out water line without the original signal and therefore cut down the bound in practical applications. The correlativity coefficient is farther improved by utilizing Dynamic Fuzzy Inference System. The primary freshness of this strategy is that the Mamdani type DFIS theoretical account is exploited in order to find a valid estimate of a quantization measure of each DWT coefficient. Furthermore, the HVS belongingss are modeled utilizing biorthogonal ripples to better water line hardiness and imperceptibility. Finally, the consequences of BPNN and DFIS methods are compared.

## Chapter Summary

This chapter presented an overview of digital image watermarking. A study is made on digital image watermarking and their restrictions are besides presented. Different spheres of watermarking are explained in the following chapter.