JPEG DCT Coefficient Extraction for Low-Latency Object Detection

Solution Overview

Problem

The efficiency of object detection in surveillance images is reduced due to limited network bandwidth or system bus congestion caused by latency in data transmission of JPEG-compressed images using Joint Photographic Experts Group (JPEG) compression.

Innovation Solution

An object detection method that reconstructs a coarse scan image using a subset of Discrete Cosine Transform (DCT) coefficients from a progressively encoded JPEG image, performing object detection on this coarse image to improve efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If full JPEG image data is transmitted and processed, then object detection accuracy is maintained, but data transmission latency increases and system bus becomes congested

Engineering Contradiction:

Improveobject detection accuracyVSAvoiddata transmission latency

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent extracts only the essential DCT coefficients (DC coefficients and low-frequency AC coefficients) from the full JPEG image data, discarding high-frequency AC coefficients that contribute minimal visual information. This extraction enables transmission of a compressed subset of data that maintains sufficient quality for object detection while significantly reducing transmission volume and latency.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the JPEG DCT coefficients into different frequency components (DC coefficients, low-frequency AC coefficients, and high-frequency AC coefficients). By processing only the first segment (DC and low-frequency AC coefficients), the system achieves a balance between reconstruction quality and transmission efficiency, resolving the contradiction between accuracy and latency.

Inventive Principle:

Principle #1Segmentation

2Loss of information

If full JPEG image data is transmitted, then complete image information is available, but system bus bandwidth is exhausted causing congestion

Engineering Contradiction:

Improveimage information completenessVSAvoidsystem bus congestion

Core Design Contradiction:

Loss of informationVSDevice complexity

Solution Approach 1:

The patent extracts only the essential DCT coefficients (DC coefficients and low-frequency AC coefficients) from the full JPEG image data, discarding high-frequency AC coefficients that contribute minimal visual information. This extraction enables transmission of a compressed subset of data that maintains sufficient quality for object detection while significantly reducing transmission volume and latency.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the parameter of data representation by transforming the full JPEG data into a reduced set of DCT coefficients. This parameter transformation maintains the essential information needed for object detection while reducing the data size to a level that prevents system bus congestion.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If more DCT coefficients are decoded to improve image quality, then reconstruction accuracy increases, but processing time and computational resources increase

Engineering Contradiction:

Improveimage reconstruction qualityVSAvoidprocessing speed

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent applies partial action by decoding only the necessary DCT coefficients (DC and low-frequency AC coefficients) rather than all coefficients. This partial decoding provides sufficient image quality for object detection purposes while significantly reducing computational load and processing time, thus resolving the contradiction between quality and speed.

Inventive Principle:

Principle #16Partial or excessive action

Solution Approach 2:

The patent segments the JPEG DCT coefficients into different frequency components (DC coefficients, low-frequency AC coefficients, and high-frequency AC coefficients). By processing only the first segment (DC and low-frequency AC coefficients), the system achieves a balance between reconstruction quality and transmission efficiency, resolving the contradiction between accuracy and latency.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20250232551A1Object detection method and object detection apparatus

Publication Date: 2025.07.17 HIMAX TECH LTD

AI summary

An object detection method and an object detection apparatus are disclosed. In the object detection method, a bitstream of a progressively encoded JPEG image is fetched from an image supply device, wherein the bitstream includes a header and a subset of a plurality of discrete cosine transform (DCT) coefficients of the JPEG image, the subset of the DCT coefficients is decoded according to information in the header to reconstruct a coarse scan image of the JPEG image; and an object detection is performed on the coarse scan image to detect an object in the JPEG image.