BIT LAYERS SPECTRAL-GROUP CODING METHOD OF INFRARED IMAGES

Improving the efficiency of remote thermal monitoring systems requires the development of coding methods for infrared (IR) images that account for their semantic characteristics. It is established that infrared images typically have high bit-depth (up to 16 bits), which significantly complicates their processing and transmission in real-time, especially under limited channel bandwidth conditions. It is substantiated that traditional coding algorithms, such as JPEG, JPEG2000, or H.265, do not provide efficient handling of IR data since they fail to consider the internal bit structure, spectral non-uniformity, and local temperature characteristics of the scene. Compression efficiency can be improved by representing an IR image as separate higher and lower bit layers with subsequent differentiated processing. The paper presents an approach that involves preliminary division of the image into 8×8 pixel segments and 4×4 mini-segments, each of which is transformed into a residual domain by subtracting its local minimum. A recursive one-dimensional Haar wavelet transform is proposed, with decomposition depth up to a single low-frequency coefficient, enabling energy concentration and isolation of multilevel frequency components. It is shown that such a representation is convenient for further spectral generalization and adaptive encoding. A functional model of group coding is developed, allowing each mini-segment to be represented by a single high-frequency code value, while all low-frequency coefficients of a segment are combined into a separate group that is also group-encoded. As a result, each image segment is represented by five code values, which significantly reduces data volume while preserving thermal semantics. The proposed method is suitable for real-time applications and systems with limited computational resources, particularly in unmanned aerial platforms and mobile thermal imaging systems.