Está en la página 1de 7

Coordinate Logic Order Statistics. A new revolutionary approach for Image & Signal processing.

Dr. Tsirikolias K.
zeniolo@windowslive.com

Prof. Mertzios B.
bgmertz@gmail.com

Abstract
A new revolutionary approach for Image & Signal processing is presented. This new approach has to do with the philosophy of finding a method for designing and executing algorithms (filters), using the minimum complexity and energy. The proposed method is based on a special formation of Coordinate Logic Filters (CLF), the Coordinate Logic Order Statistics filters (CL-OS). The CL-OS introduce a minimalistic approach for filter design followed with the most possible efficiency with respect to speed, quality, complexity and energy consumption. They may be considered as an alternative expression of Order Statistics. Their complexity O, in terms of speed and H/W interconnections, is absolutely linear. We can safely claim that the CL-OS filters are the fastest and simplest non linear digital filters, with excellent and unique performance. A number of typical image processing tasks is presented in order to prove the efficiency of CL-OS filters.

1. Introduction & general description.


Since the 80s, signal and image processing has a tremendous growth. The industrial & market sectors involve signal and image processing are many with a wide range of applications. The need for real time processing has brought the hardware design at its limits, while the innovation of new H/W and VLSI technologies is very expensive. The problem that the modern microelectronics faces, is not only the miniaturization but also the low energy consumption. That problem arises from many modern applications, for example in neuroprosthetics, the implant hardware is preferred to be supplied direct from the biological organism rather than an external (battery) source. Similar problems and provokes arises every day from the industry, for example in mobile communications, where real-time operations on video, like advanced noise cleaning, frame-level encryption & dynamic bandwidth reduction are demanded. In order to achieve those challenges, more sophisticate algorithms must be designed in order to reduce the hardware complexity and thus the energy consumption needed for their implementation!

In this presentation, an edge innovation, addressing a range of highly extensive applications in the digital imaging field, is offering the capacity to re-design existing applications with scalar performance and cost improvements. Another option is the design of new applications which would be otherwise very difficult or even not feasible to achieve with the current technological limits due to processing barrier, size, consumption or construction cost and complexity. This new approach has to do with the philosophy of finding a method for designing and executing algorithms (filters), using the minimum complexity and energy. The proposed method is based on a special formation of Coordinate Logic Filters (CLF), the Coordinate Logic Order Statistics filters (CL-OS). CL-OS introduce a minimalistic approach for filter design followed with the most possible efficiency in terms of speed, quality, complexity and energy consumption. They consider the signal as a group of elementary binary sets, they do not apply arithmetic operations, instead they apply the three fundamental logic operations of intersection (OR) , section (AND) and complement (NOT) , without carry bits. CL-OS Filters deal with reasoning that is approximate rather than fixed and exact. We could claim that the CL-OS filters, have a straight and direct relation, with the processing functions of live organisms, since their functions are based on comparisons and approximations. CL-OS decompose the signal into k binary signals that operate in parallel and achieve the desired processing, by executing only direct logic operations, among the binary values of the given signal. The philosophy of CL functioning, is the process in the bit level of the signal rather in the decimal level. It can be considered as a bottom up process. Hence a procedure that approximates the arithmetic functions by applying elementary Boolean functions, takes place. This is based in the assumption that the functionality of the MIN and MAX operators, is the same with the corresponded Boolean AND and OR operators. The last notion has a mirror effect in the designed CL-OS filters and generally in CLF. As a result of the above kind of process, the CL-OS filters reveal new signal information that help us to make a different interpretation of the original signal. From the above kind of process, new theoretical and practical results are revealed in many fields such as order theory, signal processing, cellular automata , e.t.c.. In the same time a new approach for image and signal processing is developed. The CL-OS filters, and generally the CLF, are characterized by Inherent parallelism. Real time speed. Effectiveness. Simplicity. Their complexity O in terms of speed and H/W interconnections, is absolutely linear! From the above, it is safe to claim that the CL-OS filters are the fastest and simplest non linear digital filters, while they show an excellent and unique performance. Using the CL-OS framework, a unified method for image processing applications has been developed. CLOS is an innovative approach in digital signal processing with international recognition of innovation, and one of the most promising fields in next-generation focused fields of research. The Generalized Coordinate Logic filters algorithmic framework, can assess and implement almost all the kinds of non linear filters. More specific the CL-OS may be considered as an alternative expression of the Order Statistics, OS. However CL-OS filters are much simpler and faster than the corresponded OS filters, while in the same time show the same or even better efficiency. Due their unorthodox behavior they

benefit remarkable properties, such as the ability to enhance the signal while in the same time to remove all the impulse types of noise. In other words CL-OS filters can act simultaneously as high and low pass filters, they are adaptive by their nature. Using CL-OS it is feasible to implement all the kinds of rank filters such as morphological (min-max), Median, high-pass, low-pass, band-pass, -trimmed and many others. This, practically offers an opportunity for assessing the majority of complex image processing applications in several demanding fields including medical & diagnostics, feature extraction, engineering, real-time video processing, scientific visualization, encryption, cellular automata, among a very wide range of directly related applications. The methodologies, algorithms and intense know-how, may well be applied in both Software and/or Hardware implementations. Combined with other proprietary algorithms (e.g. advanced image segmentation and object recognition), next generation information indexing and retrieval engines may be developed for image and video!

Industrial applications The main target is the realization of a single-chip based real-time image processing hardware solution, at a very low cost, making possible the easy implementation of advanced image processing solutions, either at firmware or software level. Technical superiority at a dramatically reduced cost due to reduction of logic circuitry used, with a matching reduction of energy consumption, is the claimed benefit of the approach.

2. Typical applications
1. Median filtering

Fig.1

Fig.1a

Fig.1b

Fig.1. A 24-bit RGB, image corrupted with impulse random noise. Fig.1a. Application of a 3x3 CL-Median filter. Fig.1b. Application of a 3x3 Median filter.

Fig.2

Fig.2a

Fig.2b

Fig.3

Fig.3a

Fig.3b

Fig.2. A 24-bit RGB, image corrupted with massive impulse noise. Fig.2a. Application of a 3x3 CL-Median filter. Fig.2b. Application of a 3x3 Median filter. Fig.3. An 8-bit gray level. . Fig.3a. Application of a 3x3 CL-Median filter. Fig.3b. Application of a 3x3 Median filter.

Reviewing the results As it is clear seen, the performance of the CL-Median is superior than the corresponded Median. More over, the CL- OS median filter tends to de blur the image increasing the contrast, the bigger the filter mask the bigger the contrast. On the other hand, the median filter tends to blur the image, the bigger the filter mask the bigger the blurriness. This behavior is clear observed in Fig 3, were the two filters have been applied in a clear image. In order to have a magnitude of the simplicity of the CL-OS filters an inside chip image is given. SPARTAN 3 FPGA CL-OS Median filter on chip Median filter on Chip

2. Edge extraction.

Fig.1

Fig.1a

Fig.1b

Fig. 1. Shows an image of the RGB color space palette. Fig. 1a. Shows the application of the CL-OS edge detector. Fig. 1b. Shows the application of the corresponded OS edge detector (min max).

3. Rank filtering (extremum filter).


Fig.1 Fig.1a Fig.1b

Fig 1, shows an original MRI image. Fig.1a, shows the extremum filtered image using CL-OS filters. Fig.1b, shows the extremum filtered image using OS filters

4. Morphological image processing.


Fig.1 Fig.1a Fig.1b

Fig.2

Fig.2a

Fig.2b

Fig. 1. An original 5-bit cell image. Fig. 1a. Opening using a 9x9 CLM (Coordinate Logic Morphological) filter. Fig. 1b. Opening using a 9x9 Morphological filter. Fig. 2. A preprocessed (from 8-bit to 6-bit) Lena image. Fig. 2a. Opening using a 7x7 CLM(Coordinate Logic Morphological) filter. Fig. 2b. Opening using a 15x15 Morphological filter.

También podría gustarte