# Method for compressing and decompressing digital image data

A compressing method consisting of generating a predicted value for the pixel of the digitized image to be compressed based on the values of at least two neighboring pixels is provided. After calculating the difference between the predicted value and the current pixel value, a prediction error is defined by a one-on-one mapping on the difference obtained. According to an advantageous feature of the mapping relationship, the required number of bits for representing the prediction error is less than that required for the difference per se. Afterwards, a parameter K is generated which is in proportion to the difference between the neighboring pixels used in prediction. According to the respective values of the parameter K, the next step is to encode each prediction error by assigning a variable-length Rice code to the prediction error of each pixel. If the Rice code formatted for a prediction error is longer than a predetermined length, the Rice code is disregarded and replaced by predetermined length of leading "1" bits followed by the value of current pixel.

## Latest Acer Peripherals, Inc. Patents:

- Method of producing light source compensating element and scanning appaartus applying the light source compensating element
- Image reading apparatus
- Mechanism for triggering an eject device of a disk player
- Multi-functional keyboard
- Changing the positions of lenses by way of movement of optical module

## Claims

1. A lossless compression method for a digitized image, the digitized image including a current pixel, which has a value X, and two neighboring pixels adjacent to the current pixel, the two neighboring pixels having values A and B respectively, and.DELTA. being the difference between A and B, comprising the steps:

- using a first algorithm to generate a predicted value P for the current pixel based on the values A and B;
- calculating a prediction error e' based on a one-to-one mapping of the difference e between the predicted value P and the current value X;
- generating a parameter K which is a predetermined non-constant function of the difference.DELTA.;
- encoding the prediction error e' of the current pixel by assigning a Rice code thereto based on the parameter K, to thereby obtain a Rice encoded code; and
- replacing the Rice encoded code by a predetermined code pattern followed by the current value X if the Rice encoded code obtained for the current pixel is longer than a predetermined length.

2. Method according to claim 1, wherein the predetermined code pattern is a set of consecutive "1" bits.

3. Method according to claim 1, wherein

- the difference e is a signed quantity and may have a positive value or a negative value;
- the prediction error e' is an unsigned quantity and does not include any negative values;
- a small absolute value of the difference e corresponds to a small value of the prediction error e'; and
- the number of bits required to represent the unsigned prediction error e' is less than the number of bits required to represent the signed difference e.

4. The method of claim 3, wherein the predetermined function for generating K consists essentially of:

- k=1.DELTA..ltoreq.2
- k=2 2<.DELTA..ltoreq.4
- k=3 4<.DELTA..ltoreq.8
- k=4 8<.DELTA..

5. Method according to claim 1, further comprising a subsequent decompression procedure comprising the steps of:

- determining whether the leading bits of the encoded code consists of the predetermined code pattern; and
- if the leading bits of the encoded code consist of the predetermined code pattern, outputting X.

6. Method according to claim 1, further comprising a subsequent decompression procedure comprising the steps of:

- determining whether the leading bits of the encoded code consist of the predetermined code pattern; and
- if the leading bits of the encoded code does not consist of the predetermined code pattern,
- generating a decoded predicted value P for the current pixel based on previously decoded values A and B for the neighboring pixels,
- using said predetermined function to generate a decoded parameter K from the decoded difference.DELTA. of the decoded values A and B;
- using the decoded value of the parameter K and the inverse of said Rice code to obtain a decoded prediction error e' from the encoded code;
- using a second mappling function which is the inverse of said one-to-one mapping function to obtain a decoded difference e from the decoded prediction error e'; and
- adding the decoded predicted value P to the decoded difference e to obtain the decoded value X of the current pixel.

7. The method of claim 1, wherein the predetermined function for generating K consists essentially of:

- k=1.DELTA..ltoreq.2
- k=2 2<.DELTA..ltoreq.4
- k=3 4<.DELTA..ltoreq.8
- k=4 8<.DELTA..

8. A lossless compression method for a digitized image, the digitized image including a current pixel, which has a value X, and two neighboring pixels adjacent to the current pixel, the two neighboring pixels having values A and B respectively, and.DELTA. being the difference between A and B, comprising;

- using a first algorithm to generate a predicted value P for the current pixel based on the values A and B;
- calculating a prediction error e' based on a one-to-one mapping of the difference e between the predicted value P and the current value X;
- generating a parameter K which is a predetermined non-constant function of the difference.DELTA.;
- encoding the prediction error e' of the current pixel by assigning a Rice code thereto based on the parameter K, to thereby obtain a Rice encoded code; and
- if the Rice encoded code obtained for the current pixel is longer than a predetermined length, replacing the Rice encoded code by a predetermined code pattern followed by the current value X
- determining whether the leading bits of the encoded code consists of the predetermined code pattern:
- if the leading bits of the encoded code consist of the predetermined code pattern,
- outputting X,
- determining whether the leading bits of the encoded code consist of the predetermined code pattern:
- if the leading bits of the encoded code does not consist of the predetermined code pattern,
- generating a decoded predicted value P for the current pixel based on previously decoded values A and B for the neighboring pixels,
- using said predetermined function to generate a decoded parameter K from the decoded difference.DELTA. of the decoded values A and B;
- using the decoded value of the parameter K and the inverse of said Rice code to obtain a decoded prediction error e' from the encoded code and;
- using a second mapping function which is the inverse of said one-to-one mapping function to obtain a decoded difference e from the decoded prediction error e'; and
- adding the decoded predicted value P to the decoded difference e to obtain the decoded value X of the current pixel,
- k=1.DELTA..ltoreq.2
- k=2 2<.DELTA..ltoreq.4
- k=3 4<.DELTA..ltoreq.8
- k=4 8<.DELTA..

**Referenced Cited**

**Patent History**

**Patent number**: 5751860

**Type:**Grant

**Filed**: Sep 3, 1996

**Date of Patent**: May 12, 1998

**Assignee**: Acer Peripherals, Inc.

**Inventors**: Yang-Chun Su (Taoyuang), Yu-Wen Huang (Taoyuan)

**Primary Examiner**: Edward L. Coles, Sr.

**Assistant Examiner**: Fan Lee

**Application Number**: 8/708,276

**Classifications**

**Current U.S. Class**:

**Lossless Compression (382/244);**Predictive Coding (382/238); Huffman Or Variable-length Coding (382/246); 358/2612; 358/426

**International Classification**: H04N 141;