	US 12,015,767 B2
	Intra-frame predictive coding method and system for 360-degree video and medium
Li Yu, Hubei (CN); and Zhewen Sun, Hubei (CN)
Assigned to HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY, Hubei (CN)
Filed by HUAZHONG UNIVERSITY OF SCIENCE AND TECHNOLOGY, Hubei (CN)
Filed on Aug. 9, 2023, as Appl. No. 18/447,252.
Claims priority of application No. 202211039982.6 (CN), filed on Aug. 29, 2022.
Prior Publication US 2024/0080439 A1, Mar. 7, 2024
Int. Cl. H04N 19/105 (2014.01); H04N 19/119 (2014.01); H04N 19/147 (2014.01); H04N 19/176 (2014.01); H04N 19/96 (2014.01)

CPC H04N 19/105 (2014.11) [H04N 19/119 (2014.11); H04N 19/147 (2014.11); H04N 19/176 (2014.11); H04N 19/96 (2014.11)]

10 Claims

1. An intra-frame predictive coding method for a 360-degree video, comprising:

S1: inputting a coding unit feature, a texture feature, and a 360-degree video feature of a coding unit (CU) block to be partitioned into a neural network corresponding to each partition mode when intra-frame block partition is performed on a current CU block, skipping a partition mode in which the output of the neural network is smaller than a corresponding output threshold, performing the intra-frame block partition on the CU block to be partitioned according to a partition mode that is not skipped, and obtaining a plurality of intra-frame block partition schemes of the current CU block, wherein intra-frame block partition modes comprise quadtree (QT), binary tree horizontal (BTH), binary tree vertical (BTV), tertiary tree horizontal (TTH), and tertiary tree vertical (TTV), and the CU block to be partitioned is the current CU block or a sub-CU block obtained after each intra-frame block partition,

wherein the coding unit feature comprises: several ones of a width, a height, a depth, a quadtree depth, a multi-tree depth, a block shape ration, a quantization parameter, a horizontal angle mode type, and a vertical angle mode type of a CU block,

the texture feature comprises: several ones of a variance of a pixel value, a normalized mean square error, a horizontal texture factor, a vertical texture factor, a weight between the horizontal and vertical textures, and directional complexity of the CU block, and

the 360-degree video feature comprises a latitude of the CU block and a sampling factor at each latitude;

S2: determining a length n of a rough mode decision (RMD) candidate list according to the coding unit feature and the texture feature of the sub-CU block when an intra-frame angle mode is decided for each sub-CU block in each intra-frame block partition scheme, deciding the intra-frame angle mode based on the RMD candidate list of the length n, and obtaining the intra-frame angle mode to fill pixels in each intra-frame block partition scheme,

wherein when w≥w_th, h≥h_th, var<var_th, NMSE<NMSE_th, and compT_HT_V<comp_th), the length n is 1, and

when w≥w_th&&h≥ h_th&& (ratioT_HT_V>th_H∥1/ratioT_HT_V>th_V), the length n is 2; otherwise, the length n is 3,

where w and h respectively are the width and the height the CU block, var and NMSE respectively are the variance of the pixel value and the normalized mean square error of the CU block, compT_HT_Vis the directional complexity, ratioT_HT_Vis the weight between the horizontal and vertical textures, w_th, h_th, Var_th, NMSE_th, and comp_threspectively are threshold values corresponding to w, h, var, NMSE, and compT_HT_V, th_His a threshold value of the horizontal angle mode, and th_Vis a threshold value of the vertical angle mode;

S3: calculating rate distortion optimization (RDO) loss between each intra-frame block partition scheme and the current CU block after pixel filling and performing intra-frame coding predictive coding on the current CU block based on the intra-frame block partition scheme corresponding to the minimum RDO loss; and

S4: repeatedly executing S1 to S3 until intra-frame predictive coding is performed on all CU blocks of the 360-degree video frame.