# INTRODUCTION he basic idea of compression by removal of coding redundancy is to represent the symbols with as less number of bits as possible, so that the average number of bits required to represent the data is reduced. This can be optimized by assigning shorter codes to more frequent symbols and longer codes to less frequent symbols i.e. if A(i) represents i th symbol of the symbol array A, assuming there are a total of n symbols, and let P[A(i)] represent probability of occurrence of the i th symbol, if P[A(i)]<=P(A(j)] (1) then L[A(i)]>=L(A(j)] Where L(A(i)] represents the length of code assigned to the i th symbol. Once the codes are assigned and the symbols are transmitted, the string of code words at the receiving end should be uniquely decodable. To allow unique decodability, we assign prefix codes. Prefix codes are those codes where no code assigned to a symbol is prefix to the code assigned to some other symbol [2], [4]. # II. # EXISTING METHOD The optimum coding of message symbols using the Non-Binary Huffman technique is similar to the binary Huffman Coding technique. Every iteration, the least probable D symbols are represented in the next iteration with a new imaginary symbol whose probability equals the sum of probabilities of the constituting symbols. Therefore the number of symbols Author : Email : srinivas.gitam@gmail.com, prof.prasadreddy@gmail.com, ramyaa.krothapalli@gmail.com reduces by D-1 every iteration. To assign 0,1,?.,D-1 codes to the symbols remaining in the last iteration, we need to have D symbols left. To allow this, J symbols are added up in the 1 st iteration [4]. Where J is given by the formula: J= 2 + [(N-2) % (D-1)] ---------(3) Where N is the total number of symbols in the sample. There after D least probable symbols are added up every iteration till only D symbols are left. # Illustration with an Example Consider the 18-symbol sample as shown in Table 1. The iterations and the final codes assigned are shown in Table 1(Codes are shown in the brackets). (Assuming the value of D is 5). Here the number of symbols added up in the 1 st iteration is given by J= 2+[(18-2)%(5-1)]=2 # PROPOSED METHOD The erratic nature of Non-Binary coding stems from the fact that no pattern has been proposed to chose the value of D. Different D values chosen give different results. One can take the best possible D value to reduce the number of iterations as well as the July number of symbols in each iteration. We propose a pattern to assign a value to D. # N -Square Approach to choose D value The number of symbols N determines the D value [10]. D is given by The number of symbols to be added up in the 1 st iteration is given by the formula # P= (D-1)+N% (D-1) ------------(4) where N is the total number of symbols in the sample and D is obtained from step A. Code generation and assignment to the symbols added up in the 1st iteration. In cases where the number of symbols added up in the 1 st iteration is less than or equal to the D value, the codes are obtained by concatenating 0,1,2?D-1 to the prefix. (The prefix here is the code assigned to the imaginary symbol probability which represents the symbols added up in the 1 st iteration). In cases where the number of symbols added up in the 1 st iteration exceed the D value, the prefix is concatenated with 0, 1 , 2 ?D-2 ,(D-1)0,(D-1)1?. and assigned to the symbols. # Illustration with an example Reconsider the 18-symbol example. Since the number of symbols, N=18, D is assigned 5. The number of symbols to be added up in the 1 st iteration is given by P=(5-1)+18%(5-1)=6. Then on, every iteration , 5 symbols are added up. The probability which represents the 6 symbols that are added in the 1 st iteration is assigned a code during the process, which is used as prefix to all the 6 symbols. Since the D value is 5, we can concatenate only 5 different symbols to the prefix obtained. In order to generate 6 uniquely decodable codes for the 6 symbols, we expand the obtained 5 th code by concatenating 0, 1 to it. In the example prefix 2 is obtained for the 6 symbols. There 20,21,22,23,240,241 are the final 6 codes assigned to the symbols. It is evident that the number of iterations are reduced in the proposed method without effecting the average length of the code. In cases where a large number of symbols is involved, complexity is reduced by a greater degree since the number of symbols reduced per iteration is maximized. # Original # DECODING The codes pertaining to the symbols are sent across from the sender to the receiving end in the form of strings of digits ranging from 0 to D-1. Since the codes assigned are prefix codes[2], the string can be uniquely decodable. A single pass over the string can allow us to determine the unique symbols which correspond to the codes. V. # CONCLUSION Experiments show that the proposed method allows assigning codes to the symbols using lesser number of iterations than those employed in the original Non-Binary Huffman Coding. In case of samples where the number of symbols are too many, the number of iterations are considerably reduced thereby reducing the complexity involved. 3 © 2011 Global Journals Inc. (US) © 2011 Global Journals Inc. (US) July July © 2011 Global Journals Inc. (US) July * SJayaram SEsakkirajan TVeerakumar Digital Image Processing Tata Mc grawHill Publications * Approach for Reduced Complexity Non-Binary Encoding Global Journal of Computer Science and Technology Volume XI Issue XI Version I 2011 2. Rafel c N-Square ;An RichardEGonzalez Woods Digital Image Processing 2009 Second Edition, 6Th Impression * SAnnadurai RShanmughalakshmi Fundamental of Digital Image processing 3rd * * A Method for the Construction of Minimum-Redundancy Codes David A Huffman Associate Ire 1952 * An Overview of Image Compression Approaches CebrailTaskin Kür?atSerdar Sarikoz 2008 ICDT * GLu Advances in digital image compression techniques 1993 * Huffman Coding Method based on Number Character BeiChen Hong-CaiZhang CaoWen-Lun Jian-HuFeng ICMLC 19-22 Aug-2007 IEEE * Direct Huffman Coding and Decoding using the table of Code-Lengths RezaHashemian ITCC'03 IEEE * Balance of 0,1 Bits for Huffman and Reversible Variable-Length Coding Jia-YuLin YingLiu Ke-ChuYi IEEE Transactions on Communications 52 3 March 2004 * N-Square Approach for the Erection of Redundancy Codes GSrinivas International Journal of Engineering Science and Technology 2 3 2010