CN105657399A - 3D medical video transmission method in wireless network environment - Google Patents

Abstract

The invention discloses a 3D medical video transmission method in a wireless network environment. An end to end video distortion in a mode of a current coding QP (quantization parameter), a current authentication Hash quantity and current MCS (modulation and channel coding schemes) is obtained through recursive computation; the end to end video distortions under combination conditions of all coding QPs, authentication Hash quantities and MCS are computed through traversal; the QP, the authentication Hash quantity and the current MCS under a least distortion condition are selected for coding and authenticating an application layer, configuring the physical layer MCS and transmitting current data packets. Through the mode, the strengths of different parameters in authentication and fault-tolerant aspects can be configured and coordinated systematically; and the quality of the 3D medical video can be maximized in the current network state.

Description

3D medical video transmission method in wireless network environment

Technical Field

The invention relates to the field of video coding and wireless transmission, in particular to a 3D medical video transmission method in a wireless network environment.

Background

With the rapid development of electronic information and multimedia communication technologies, modern medical technologies gradually move to a digital distributed application mode. The modern development of the medical health industry is greatly promoted by the electronic health technology generated by combining the information technology and the biotechnology. The core problem of the electronic health technology is to remotely process and monitor the health information of the patient. The combination of video processing and wireless communication technology breaks through the regional limitation of medical level difference, and a remote medical system is produced. Most of the traditional telemedicine systems are based on 2D video communication systems to transmit and interact medical information. In recent years, with the rapid development of technologies such as digital imaging, 3D vision, and 3D display, the digital 3D video technology has become a research hotspot in academia and industry at present. Compared with a 2D video, the 3D video can enable the foreground image and the background image to be distinguished more obviously, and has clearer boundary outline and finer texture patterns, so that better imaging quality and more vivid and natural 3D perception effect can be presented to people, people can see the three-dimensional nature of a natural scene again, and the people can feel visual experience which is completely different from that of a common plane picture.

To provide more accurate third-dimensional information, 3D video may be applied to telemedicine systems. The 3D video system can enable a doctor to obtain depth information data which cannot be captured from the traditional plane display, can read image information in all directions, provides richer and more accurate image data for clinical diagnosis, and greatly reduces missed diagnosis of a focus, thereby improving diagnosis quality.

Currently, 3D medical video communication is mostly based on wireless networks, which can cover remote areas. Since wireless communication technology is more vulnerable to intrusion than wired communication, authentication technology in the medical video communication process is very critical to protecting the privacy of patients. The traditional authentication signature technology facing data flow signs each data packet, and the cost such as complexity and the like is high.

In the data packet encoding process, each encoded slice (slice) is packed as a basic packing unit. And the code rate of each video data packet is basically determined by the quantization parameter of the coding. Therefore, in real-time coding, the Quantization Parameter (QP) of the coded slice can be determined in real time according to the channel bandwidth, and the selection of different QPs will result in different levels of video distortion, i.e. different levels of video quality. Each encoded video data packet is transmitted as a basic application layer unit to be transmitted to other layers of the network.

In order to ensure the security and easy copyright management of video data packets, video transmission usually requires an authentication technique for encryption or authentication processing. In the existing video transmission, a multi-path hash chain authentication mode, that is, a hash authentication processing mode based on a same source multi-chain, may be used to authenticate a group of pictures (GOP). The packets of each frame may be selected by parent nodes (the upper-level indices of the hash chain) that hash the corresponding position packets within the already decoded frame within the GOP. There may be several hash chain path selection schemes (i.e. parent node selection) for each packet, and each scheme estimates the authentication probability of the current authentication scheme of the current packet according to the authentication of the previous node on the path. While calculating the number of bits spent for the authentication.

Assuming that each packet can have a maximum of N hash parents, the number of bits consumed by each hash is b, and the probability of successful authentication isThe probability of channel transmission packet loss is assumed to be p. Each packet may have multiple hash links such that the probability of successful authentication for each packet is different, assuming that N packets as a group have a signature. To reduce authentication and decoding delays, we assume that the current packet can only select as the parent node of the hash the packet on which the current packet depends that must be decoded earlier than the current packet. The ith (i) th node is present, assuming there are M hash parent nodes available<M) probability of successful authentication of data packets is

If it is currently the ith (i)<M) number of hash parent nodes selected by data packets is M (M)<M) and the nth parent node is a distance d from the current packet_nThen the probability of successful authentication of the current data packet is revised to

And the number of bits spent for hash authentication is m · b. Assuming that the length of the original data packet is S and the length of the packet after adding the hash authentication data is S + m · b, the number of resource blocks that each data packet can be divided into in the physical layer and the corresponding modulation and channel coding (MCS) may change according to the quality of the channel. Therefore, as the number of hash parent nodes increases, the number of resource blocks corresponding to each packet also increases, and the packet loss probability of each corresponding data packet also increases, so that the hash number of each data packet directly affects the authentication probability and the packet loss probability of data.

In addition to source and channel distortion during video transmission, video packets that fail authentication will also cause video distortion. This is mainly because the current data packet is distorted due to the fact that the current data packet cannot be authenticated because of the packet loss of the data packet on which the hash authentication depends. On the other hand, the authentication bit data occupies bandwidth data, which may cause the loss of the quality of the information source due to the reduction of the information source data, that is, a certain distortion is generated.

The stream-level authentication based on the hash chain (hashchain) signature is simple. In order to improve the fault tolerance of the authentication data packet, a homologous multi-chain authentication mode is generally adopted. But the homologous multiple chain authentication mode greatly increases redundancy of transmitted data. How to effectively select the number of authenticated hash paths directly affects the efficiency of video data transmission and the distortion of a receiving end video. Because the wireless network physical layer supports a certain modulation and channel coding mode, the data stream can be protected to a certain extent, but the selection of the physical layer modulation and channel coding mode also directly influences the packet loss probability and video distortion of the data packet. The traditional authentication method does not select authentication hash from the perspective of the overall optimization of the system, so that the optimization of the wireless communication quality of the 3D medical video cannot be guaranteed under the dynamic channel condition.

Disclosure of Invention

The technical problems to be solved by the invention are as follows: the 3D medical video transmission method under the wireless network environment with good video transmission quality is provided.

In order to solve the technical problems, the invention is realized by the following technical scheme: A3D medical video transmission method under a wireless network environment is characterized in that: the method comprises the following specific steps:

1) according to the selection of the number of the wireless 3D medical video hash capable of fault-tolerant authentication, the selection of an application layer video coding quantization parameter and the selection of an LTE downlink physical layer modulation and coding mode, the following optimal parameters are comprehensively established by minimizing end-to-end 3D video distortion:

$(m_{hash}{\left(d_0,...d_{m-1}\right)}^{opt},{\mathrm{QP}}^{opt},{\mathrm{MCS}}^{opt})=\underset{m\&Element;Q,QP\&Element;Q,MCS\&Element;Z}{\arg min}D(m,QP,MCS)$

subjecttot<T_max

wherein m is_hash(d₀,...d_m-1)^optIs the optimal Hash father node of the current data packet, H, Q, and Z respectively represent the candidate Hash father node set, QP set and MCS mode set, QP is the quantization parameter of the coding strip, MCS is the physical layer modulation and coding mode, T_maxThe maximum time delay limit of data packet transmission, T is the transmission time delay of the data packet, and can be obtained approximately by dividing the size of the data packet by the transmission rate of a physical layer;

2) setting parameters such as initial QP and MCS and the like: setting QP to minQP, minD to 0, maxMCS to 15, and maxQP to 45;

3) aiming at a 3D medical video acquired by a binocular camera, coding a coding structure with inter-view prediction according to multi-view video coding, and coding by using QP as a quantization parameter;

4) setting the data volume of a Hash father node to be 0, namely, m is 0, obtaining the maximum possible Hash node quantity of the current data packet to be maxH, and obtaining the maximum possible Hash node quantity of the current data packet according to a formulaCalculating the probability of successful authentication of the current ith data packet, wherein the current ith (i) th data packet<maxH) number of hash parent node selected by packet is m (m)<maxH) and the tth parent node is a distance d from the current packet_tThe probability of successful authentication isThe probability of packet loss of the current data transmission application layer is assumed to be rho_i；

5) Setting the coding and modulation mode MCS of the physical layer channel to be 1; for each modulation and coding mode h, calculating the equivalent signal-to-noise ratio ${\mathrm{\&gamma;}}_{mieff}\left(h\right)=\mathrm{\&kappa;}\left(h\right){\&lsqb;J^{-1}\left(\frac{1}{N_{sb}}\underset{j=1}{\overset{N_{sb}}{\&Sigma;}}J\right(\sqrt{\frac{{\mathrm{\&gamma;}}_j}{\mathrm{\&kappa;}\left(h\right)}}))\&rsqb;}^2;$

Wherein N is_sbRepresenting the number of sub-carriers, gamma_jIs the signal to interference plus noise ratio of the jth sub-carrier, k (h) is the correction factor in modulation and coding mode h, J (x) and J^-1(y) is calculated as follows:

$J\left(x\right)\&ap;\left\{\begin{array}{cc}-0.04210610x^3+0.209252x^2-0.00640081x,& 0<x<1.6363\\ 1-\exp (0.00181491x^3-0.142675x^2-0.08220540x+0.0549608),& x\&GreaterEqual;1.6363\end{array}\right.$

$J^{-1}\left(y\right)\&ap;\left\{\begin{array}{cc}1.09542y^2+0.214217y+2.33727\sqrt{y},& 0<y<0.3646\\ -0.706692log(-0.386013(y-1\left)\right)+1.75017y,& 0.3646\&le;y\&le;1\end{array}\right.;$

6) calculating the physical layer resource block loss rate and the application layer packet loss rate of the current data packet according to the signal-to-noise ratio (SNR):

the block loss probability for a resource block with the selected modulation and coding mode h is calculated as follows:

where erfc (-) is a complementary error function, and b (h) and c (h) can be obtained in advance by means of fitting calculations;

packet loss probability of ith packet in nth frame of application layer video code streamWherein, B_numThe physical layer resource occupied by the current data packet is fast;

7) calculating distortion according to the coding structure, wherein the packet loss rate in the distortion calculation comprises the combination of the authentication probability and the packet loss probability, namely the probability that the current packet is successfully authenticated and does not lose the packet is

8) Calculating end-to-end distortion D in the current data packet transmission process:

$\begin{array}{c}D=\frac{1}{S}\underset{i=1}{\overset{S}{\&Sigma;}}E\&lsqb;{(f_n^i-{\tilde{f}}_n^i)}^2\\ =\frac{1}{S}\underset{i=1}{\overset{S}{\&Sigma;}}{\left(f_n^i\right)}^2-2\&CenterDot;f_n^i\&CenterDot;E\&lsqb;\left({\tilde{f}}_n^i\right)\&rsqb;+E\&lsqb;{\left({\tilde{f}}_n^i\right)}^2\&rsqb;\end{array};$

wherein, S is the number of the image contained in the current data packet in the current nth frame;

if the current frame belongs to the left viewpoint, a traditional distortion calculation method is adopted. The distortion is calculated by using a method in which the pixel is expected to recur. Distortion calculation for intra-frame predictive coding block

$E\&lsqb;{\tilde{f}}_{n,l}^i\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;\left({\tilde{f}}_{n,l}^i\right)+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^k\&rsqb;+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^i\&rsqb;;$

Image number distortion calculation for inter-predicted coding blocks

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,l}^i\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;({\hat{e}}_{n,l}^i+E\&lsqb;{\tilde{f}}_{n-1,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^i\&rsqb;\end{array};$

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,l}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;\{{\left({\hat{e}}_{n,l}^i\right)}^2+2{\hat{e}}_{n,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n-1,l}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,l}^j\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,l}^i\right)}^2\&rsqb;\end{array};$

And if the current frame belongs to the right viewpoint, distortion calculation is carried out according to different prediction relations: for theThe calculation of (a) can be classified into the following cases:

(1) the current pixel is located in the coding block by adopting intra-frame prediction,

$E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\left({\tilde{f}}_{n,r}^i\right)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;;$

(2) the coding block where the current pixel is located adopts inter prediction,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\end{array};$

(3) the coding block where the current pixel is located adopts inter-view prediction,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\end{array};$

(4) the weighted prediction of the inter prediction and the inter prediction employed by the coding block in which the current pixel is located,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^o\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\end{array};$

(1) the current pixel is in the coding block by adopting intra-frame prediction

(2) The coding block where the current pixel is located adopts inter prediction,

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

(3) the current pixel is in the coding block using inter-view prediction

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

(4) The coding block where the current pixel is located adopts weighted prediction between views and frames,

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&CenterDot;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\\ \&ap;{\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&CenterDot;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\\ \&ap;{\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

wherein,is the value of the ith pixel in the nth frame,is the ith pixel value, E [. in the nth frame at the decoder end]In order to be a function of the expectation,for the ith frame of the left view at the decoder endThe value of each of the pixels is calculated,for the ith pixel value in the (n-1) th frame of the left view at the decoder side,is the error concealment value of the ith pixel in the (n-1) th frame of the left view at the decoder end,for the reconstructed value of the ith pixel in the nth frame at the encoder end,is a pixelThe probability of the authentication being successful is determined,is a pixelThe probability of the authentication being successful is determined,is a pixelThe coded prediction residual of (2) is,is a pixelThe coded prediction value of (a) is,for the ith pixel value of the nth frame of the right view at the decoder end,for the ith pixel value in the n-1 th frame of the right view at the decoder side,for the error concealment value of the ith pixel of the (n-1) th frame of the right view at the decoder end,is the reconstructed value of the ith pixel in the nth frame of the right view at the encoder end,is a pixelThe probability of the authentication being successful is determined,is a pixelThe probability of the authentication being successful is determined,is a pixelThe residual values of the encoding of (1),is a pixelThe coded prediction value of (a) is,is a pixelA coding prediction value in a left view; the data can be directly obtained in the encoding process;andobtained by step 7;

9) if minD > D, minD is equal to D, and m _ opt is equal to m, MCS _ opt is equal to MCS, QP _ opt is equal to QP, and MCS is equal to MCS + 1;

10) if the MCS is less than or equal to maxMCS, returning to the step 6;

11) if m is less than or equal to maxH, returning to the step 4;

12) QP + 1; if the QP is less than or equal to maxQP, returning to the step 3;

13) and outputting m _ opt, MCS _ opt and QP _ opt, and using the parameters to encode the current 3D video data packet, configure the number of Hash authentication father nodes, select a physical layer MCS and perform wireless transmission.

By adopting the structure, the invention has the advantages that: and selecting one number of the Hash chains according to the number of the candidate Hash chains, and then estimating the corresponding probability of successful authentication. Then, the number of MCS modes of the physical layer candidate is determined, one MCS is selected, an equivalent SNR is determined, and then the block loss rate corresponding to the current MCS is estimated. According to the quantization parameter of the video coding, the alternative video coding code rate of the current data packet is obtained, a Quantization Parameter (QP) is selected, and the current packet loss probability is estimated. And obtaining the current coding QP, the current authentication Hash quantity and the end-to-end video distortion under the current MCS mode through recursive calculation. And performing application layer coding and authentication by traversing and calculating end-to-end distortion under the combination condition of all QP, authentication Hash quantity and MCS, selecting the QP, the authentication Hash quantity and the MCS under the condition of minimum distortion, configuring the MCS of a physical layer, and transmitting the current data packet. The method can configure and coordinate the intensity of different parameters on the aspects of authentication and fault tolerance, and can maximize the quality of the 3D medical video under the current network state.

Drawings

Fig. 1 is an encoding structure of inter-view prediction for a binocular acquired 3D medical video.

Fig. 2 is a hash-based authentication mechanism.

Fig. 3 is a 3D medical video authentication and fault tolerant transmission module framework.

Table 1 shows MCS modes of downlink candidates in a wireless environment.

As shown, I-Intra coded frames, P-uni-directional predictive coded frames, B-bi-directional predictive coded frames.

Detailed Description

The invention is further described below with reference to the accompanying drawings.

In the description of the present invention, it is to be understood that the terms "upper", "lower", "inside", "outside", and the like indicate orientations or positional relationships based on the positional relationships illustrated in the drawings, and are only for convenience in describing the present invention or simplifying the description, but do not indicate that a specific orientation is necessary.

The Intra-coded frames are Intra-prediction coded.

TABLE 1

As shown in fig. 1, fig. 2, fig. 3 and table 1, a 3D medical video transmission method in a wireless network environment is characterized in that: the method comprises the following specific steps:

1) according to the selection of the number of the wireless 3D medical video hash capable of fault-tolerant authentication, the selection of an application layer video coding quantization parameter and the selection of an LTE downlink physical layer modulation and coding mode, the following optimal parameters are comprehensively established by minimizing end-to-end 3D video distortion:

$(m_{hash}{\left(d_0,...d_{m-1}\right)}^{opt},{\mathrm{QP}}^{opt},{\mathrm{MCS}}^{opt})=\underset{m\&Element;Q,QP\&Element;Q,MCS\&Element;Z}{\arg min}D(m,QP,MCS)$

subjecttot<T_max

wherein m is_hash(d₀,...d_m-1)^optIs the optimal Hash father node of the current data packet, H, Q, and Z respectively represent the candidate Hash father node set, QP set and MCS mode set, QP is the quantization parameter of the coding strip, MCS is the physical layer modulation and coding mode, T_maxThe maximum time delay limit of data packet transmission, T is the transmission time delay of the data packet, and can be obtained approximately by dividing the size of the data packet by the transmission rate of a physical layer;

2) setting parameters such as initial QP and MCS and the like: setting QP to minQP, minD to 0, maxMCS to 15, and maxQP to 45;

3) aiming at a 3D medical video acquired by a binocular camera, coding a coding structure with inter-view prediction according to multi-view video coding, and coding by using QP as a quantization parameter;

4) setting the data volume of a Hash father node to be 0, namely, m is 0, obtaining the maximum possible Hash node quantity of the current data packet to be maxH, and obtaining the maximum possible Hash node quantity of the current data packet according to a formulaCalculating the probability of successful authentication of the current ith data packet, wherein the current ith (i) th data packet<maxH) number of hash parent node selected by packet is m (m)<maxH) and the tth parent node is a distance d from the current packet_tThe probability of successful authentication isThe probability of packet loss of the current data transmission application layer is assumed to be rho_i；

5) Setting the coding and modulation mode MCS of the physical layer channel to be 1; for each modulation and coding mode h, calculating the equivalent signal-to-noise ratio ${\mathrm{\&gamma;}}_{mieff}\left(h\right)=\mathrm{\&kappa;}\left(h\right){\&lsqb;J^{-1}\left(\frac{1}{N_{sb}}\underset{j=1}{\overset{N_{sb}}{\&Sigma;}}J\right(\sqrt{\frac{{\mathrm{\&gamma;}}_j}{\mathrm{\&kappa;}\left(h\right)}}))\&rsqb;}^2;$

Wherein N is_sbRepresentational sonNumber of carriers, gamma_jIs the signal to interference plus noise ratio of the jth sub-carrier, k (h) is the correction factor in modulation and coding mode h, J (x) and J^-1(y) is calculated as follows:

$J\left(x\right)\&ap;\left\{\begin{array}{cc}-0.04210610x^3+0.209252x^2-0.00640081x,& 0<x<1.6363\\ 1-\exp (0.00181491x^3-0.142675x^2-0.08220540x+0.0549608),& x\&GreaterEqual;1.6363\end{array}\right.$

$J^{-1}\left(y\right)\&ap;\left\{\begin{array}{cc}1.09542y^2+0.214217y+2.33727\sqrt{y},& 0<y<0.3646\\ -0.706692log(-0.386013(y-1\left)\right)+1.75017y,& 0.3646\&le;y\&le;1\end{array}\right.;$

6) calculating the physical layer resource block loss rate and the application layer packet loss rate of the current data packet according to the signal-to-noise ratio (SNR):

the block loss probability for a resource block with the selected modulation and coding mode h is calculated as follows:

where erfc (-) is a complementary error function, and b (h) and c (h) can be obtained in advance by means of fitting calculations;

packet loss probability of ith packet in nth frame of application layer video code streamWherein, B_numThe physical layer resource occupied by the current data packet is fast;

7) calculating distortion according to the coding structure, wherein the packet loss rate in the distortion calculation comprises the combination of the authentication probability and the packet loss probability, namely the probability that the current packet is successfully authenticated and does not lose the packet is

8) Calculating end-to-end distortion D in the current data packet transmission process:

$\begin{array}{c}D=\frac{1}{S}\underset{i=1}{\overset{S}{\&Sigma;}}E\&lsqb;{(f_n^i-{\tilde{f}}_n^i)}^2\\ =\frac{1}{S}\underset{i=1}{\overset{S}{\&Sigma;}}{\left(f_n^i\right)}^2-2\&CenterDot;f_n^i\&CenterDot;E\&lsqb;\left({\tilde{f}}_n^i\right)\&rsqb;+E\&lsqb;{\left({\tilde{f}}_n^i\right)}^2\&rsqb;\end{array};$

wherein, S is the number of the image contained in the current data packet in the current nth frame;

if the current frame belongs to the left viewpoint, a traditional distortion calculation method is adopted. The distortion is calculated by using a method in which the pixel is expected to recur. Distortion calculation for intra-frame predictive coding block

$E\&lsqb;{\tilde{f}}_{n,l}^i\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;\left({\tilde{f}}_{n,l}^i\right)+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^k\&rsqb;+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^i\&rsqb;;$

Image number distortion calculation for inter-predicted coding blocks

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,l}^i\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;({\hat{e}}_{n,l}^i+E\&lsqb;{\tilde{f}}_{n-1,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^i\&rsqb;\end{array};$

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,l}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;\{{\left({\hat{e}}_{n,l}^i\right)}^2+2{\hat{e}}_{n,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n-1,l}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,l}^j\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,l}^i\right)}^2\&rsqb;\end{array};$

And if the current frame belongs to the right viewpoint, distortion calculation is carried out according to different prediction relations: for theThe calculation of (a) can be classified into the following cases:

(1) the current pixel is located in the coding block by adopting intra-frame prediction,

$E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\left({\tilde{f}}_{n,r}^i\right)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;;$

(2) the coding block where the current pixel is located adopts inter prediction,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\end{array};$

(3) the coding block where the current pixel is located adopts inter-view prediction,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\end{array};$

(4) the weighted prediction of the inter prediction and the inter prediction employed by the coding block in which the current pixel is located,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^o\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\end{array};$

(1) the current pixel is in the coding block by adopting intra-frame prediction

(2) The coding block where the current pixel is located adopts inter prediction,

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

(3) the current pixel is in the coding block using inter-view prediction

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

(4) The coding block where the current pixel is located adopts weighted prediction between views and frames,

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&CenterDot;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\\ \&ap;{\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&CenterDot;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\\ \&ap;{\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

wherein,is the value of the ith pixel in the nth frame,is the ith pixel value, E [. in the nth frame at the decoder end]In order to be a function of the expectation,for the ith pixel in the nth frame of the left view at the decoder endThe value of the one or more of,for the ith pixel value in the (n-1) th frame of the left view at the decoder side,is the error concealment value of the ith pixel in the (n-1) th frame of the left view at the decoder end,for the reconstructed value of the ith pixel in the nth frame at the encoder end,is a pixelThe probability of the authentication being successful is determined,is a pixelThe probability of the authentication being successful is determined,is a pixelThe coded prediction residual of (2) is,is a pixelThe coded prediction value of (a) is,for the ith pixel value of the nth frame of the right view at the decoder end,for the ith pixel value in the n-1 th frame of the right view at the decoder side,for the error concealment value of the ith pixel of the (n-1) th frame of the right view at the decoder end,is the reconstructed value of the ith pixel in the nth frame of the right view at the encoder end,is a pixelThe probability of the authentication being successful is determined,is a pixelThe probability of the authentication being successful is determined,is a pixelThe residual values of the encoding of (1),is a pixelThe coded prediction value of (a) is,is a pixelA coding prediction value in a left view; the data can be directly obtained in the encoding process;andobtained by step 7;

9) if minD > D, minD is equal to D, and m _ opt is equal to m, MCS _ opt is equal to MCS, QP _ opt is equal to QP, and MCS is equal to MCS + 1;

10) if the MCS is less than or equal to maxMCS, returning to the step 6;

11) if m is less than or equal to maxH, returning to the step 4;

12) QP + 1; if the QP is less than or equal to maxQP, returning to the step 3;

13) and outputting m _ opt, MCS _ opt and QP _ opt, and using the parameters to encode the current 3D video data packet, configure the number of Hash authentication father nodes, select a physical layer MCS and perform wireless transmission.

And selecting one number of the Hash chains according to the number of the candidate Hash chains, and then estimating the corresponding probability of successful authentication. Then, the number of MCS modes of the physical layer candidate is determined, one MCS is selected, an equivalent SNR is determined, and then the block loss rate corresponding to the current MCS is estimated. According to the quantization parameter of the video coding, the alternative video coding code rate of the current data packet is obtained, a Quantization Parameter (QP) is selected, and the current packet loss probability is estimated. And obtaining the current coding QP, the current authentication Hash quantity and the end-to-end video distortion under the current MCS mode through recursive calculation. And performing application layer coding and authentication by traversing and calculating end-to-end distortion under the combination condition of all QP, authentication Hash quantity and MCS, selecting the QP, the authentication Hash quantity and the MCS under the condition of minimum distortion, configuring the MCS of a physical layer, and transmitting the current data packet. The method can configure and coordinate the intensity of different parameters on the aspects of authentication and fault tolerance, and can maximize the quality of the 3D medical video under the current network state.

The present invention and its embodiments have been described above, and the description is not intended to be limiting, and the drawings are only one embodiment of the present invention, and the actual structure is not limited thereto. In summary, those skilled in the art should appreciate that they can readily use the disclosed conception and specific embodiments as a basis for designing or modifying other structures for carrying out the same purposes of the present invention without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (1)

1. A3D medical video transmission method under a wireless network environment is characterized in that: the method comprises the following specific steps:

1) according to the selection of the number of the wireless 3D medical video hash capable of fault-tolerant authentication, the selection of an application layer video coding quantization parameter and the selection of an LTE downlink physical layer modulation and coding mode, the following optimal parameters are comprehensively established by minimizing end-to-end 3D video distortion:

$(m_{hash}{\left(d_0,\mathrm{...}{d}_{m-1}\right)}^{opt},{\mathrm{QP}}^{opt},{\mathrm{MCS}}^{opt})=\underset{m\&Element;H,QP\&Element;Q,MCS\&Element;Z}{\arg \min }D(m,QP,MCS)$

subjecttot<T_max

wherein m is_hash(d₀,...d_m-1)^optIs the optimal Hash father node of the current data packet, H, Q, and Z respectively represent the candidate Hash father node set, QP set and MCS mode set, QP is the quantization parameter of the coding strip, MCS is the physical layer modulation and coding mode, T_maxThe maximum delay limit of a data packet transmission, T, index the transmission delay of the data packet, can be obtained by dividing the size of the data packet by the physical layer transmissionThe output rate is approximately obtained;

2) setting parameters such as initial QP and MCS and the like: setting QP to minQP, minD to 0, maxMCS to 15, and maxQP to 45;

3) aiming at a 3D medical video acquired by a binocular camera, coding a coding structure with inter-view prediction according to multi-view video coding, and coding by using QP as a quantization parameter;

4) setting the data volume of a Hash father node to be 0, namely, m is 0, obtaining the maximum possible Hash node quantity of the current data packet to be maxH, and obtaining the maximum possible Hash node quantity of the current data packet according to a formulaCalculating the probability of successful authentication of the current ith data packet, wherein the current ith (i) th data packet<maxH) number of hash parent node selected by packet is m (m)<maxH) and the tth parent node is a distance d from the current packet_tThe probability of successful authentication isThe probability of packet loss of the current data transmission application layer is assumed to be rho_i；

5) Setting the coding and modulation mode MCS of the physical layer channel to be 1; for each modulation and coding mode h, calculating the equivalent signal-to-noise ratio ${\mathrm{\&gamma;}}_{mieff}\left(h\right)=\mathrm{\&kappa;}\left(h\right){\&lsqb;J^{-1}\left(\frac{1}{N_{sb}}\underset{j=1}{\overset{N_{sb}}{\&Sigma;}}J\right(\sqrt{\frac{{\mathrm{\&gamma;}}_j}{\mathrm{\&kappa;}\left(h\right)}}\left)\right)\&rsqb;}^2;$

Wherein N is_sbRepresenting the number of sub-carriers, gamma_jIs the signal to interference plus noise ratio of the jth sub-carrier, k (h) is the correction factor in modulation and coding mode h, J (x) and J^-1(y) is calculated as follows:

$J\left(x\right)\&ap;\left\{\begin{array}{cc}-0.04210610x^3+0.209252x^2-0.00640081x,& 0<x<1.6363\\ 1-\exp (0.00181491x^3-0.142675x^2-0.08220540x+0.0549608),& x\&GreaterEqual;1.6363\end{array}\right.$

$J^{-1}\left(y\right)\&ap;\left\{\begin{array}{cc}1.09542y^2+0.214217y+2.33727\sqrt{y},& 0<y<0.3646\\ -0.706692log(-0.386013(y-1\left)\right)+1.75017y,& 0.3646\&le;y\&le;1\end{array}\right.;$

6) calculating the physical layer resource block loss rate and the application layer packet loss rate of the current data packet according to the signal-to-noise ratio (SNR):

the block loss probability for a resource block with the selected modulation and coding mode h is calculated as follows:

where erfc (-) is a complementary error function, and b (h) and c (h) can be obtained in advance by means of fitting calculations;

packet loss probability of ith packet in nth frame of application layer video code streamWherein, B_numThe physical layer resource occupied by the current data packet is fast;

7) calculating distortion from coding structure, distortion meterThe packet loss rate in the calculation includes the combination of the authentication probability and the packet loss probability, that is, the probability that the current packet is successfully authenticated and does not lose packet is

8) Calculating end-to-end distortion D in the current data packet transmission process:

$\begin{array}{c}D=E\&lsqb;{(f_n^i-{\tilde{f}}_f^i)}^2\\ ={\left(f_n^i\right)}^2-2\&CenterDot;f_n^i\&CenterDot;E\&lsqb;\left({\tilde{f}}_n^i\right)+E\&lsqb;{\left({\tilde{f}}_n^i\right)}^2\&rsqb;\end{array};$

wherein, S is the number of the image contained in the current data packet in the current nth frame;

if the current frame belongs to the left viewpoint, a traditional distortion calculation method is adopted. The distortion is calculated by using a method in which the pixel is expected to recur. Distortion calculation for intra-frame predictive coding block

$E\&lsqb;{\tilde{f}}_{n,l}^i\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;\left({\tilde{f}}_{n,l}^i\right)+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^k\&rsqb;+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^i\&rsqb;;$

Image number distortion calculation for inter-predicted coding blocks

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,l}^i\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;({\hat{e}}_{n,l}^i+E\&lsqb;{\tilde{f}}_{n-1,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^i\&rsqb;\end{array};$

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,l}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,l}^i\&CenterDot;\{{\left({\hat{e}}_{n,l}^i\right)}^2+2{\hat{e}}_{n,l}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,l}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n-1,l}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,l}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,l}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,l}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,l}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,l}^i\right)}^2\&rsqb;\end{array};$

And if the current frame belongs to the right viewpoint, distortion calculation is carried out according to different prediction relations: for theThe calculation of (a) can be classified into the following cases:

(1) the current pixel is coded using intra prediction,

$E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\left({\tilde{f}}_{n,r}^i\right)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;;$

(2) the coding block where the current pixel is located adopts inter prediction,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\end{array};$

(3) the coding block where the current pixel is located adopts inter-view prediction,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^i\&rsqb;\end{array};$

(4) the weighted prediction of the inter prediction and the inter prediction employed by the coding block in which the current pixel is located,

$\begin{array}{c}E\&lsqb;{\tilde{f}}_{n,r}^i\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;({\hat{e}}_{n,r}^i+w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^o\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)+(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^k\&rsqb;\end{array};$

(1) the current pixel is in the coding block by adopting intra-frame prediction

(2) The coding block where the current pixel is located adopts inter prediction,

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

(3) the current pixel is in the coding block using inter-view prediction

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;+E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

(4) The coding block where the current pixel is located adopts weighted prediction between views and frames,

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&CenterDot;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\\ \&ap;{\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

$\begin{array}{c}E\&lsqb;{\left({\tilde{f}}_{n,r}^i\right)}^2\&rsqb;={\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&CenterDot;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\\ \&ap;{\mathrm{\&theta;}}_{n,r}^i\&CenterDot;\{{\left({\hat{e}}_{n,r}^i\right)}^2+2{\hat{e}}_{n,r}^i\&CenterDot;(w_1\&CenterDot;E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;+w_2\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;)\\ +{w_1}^2\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^j\right)}^2\&rsqb;+{w_2}^{2}E\&lsqb;{\left({\tilde{f}}_{n,l}^j\right)}^2\&rsqb;+2w_1{w}_{2}E\&lsqb;{\tilde{f}}_{n-1,r}^j\&rsqb;\&CenterDot;E\&lsqb;{\tilde{f}}_{n,l}^j\&rsqb;\}\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;{\mathrm{\&theta;}}_{n-1,r}^i\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^k\right)}^2\&rsqb;\\ +(1-{\mathrm{\&theta;}}_{n,r}^i)\&CenterDot;(1-{\mathrm{\&theta;}}_{n-1,r}^i)\&CenterDot;E\&lsqb;{\left({\tilde{f}}_{n-1,r}^i\right)}^2\&rsqb;\end{array}$

wherein,is the value of the ith pixel in the nth frame,is the ith pixel value, E [. in the nth frame at the decoder end]In order to be a function of the expectation,for the ith pixel value in the nth frame of the left view at the decoder side,for the ith pixel value in the (n-1) th frame of the left view at the decoder side,is the error concealment value of the ith pixel in the (n-1) th frame of the left view at the decoder end,for the reconstructed value of the ith pixel in the nth frame at the encoder end,is a pixelThe probability of the authentication being successful is determined,is a pixelThe probability of the authentication being successful is determined,is a pixelThe coded prediction residual of (2) is,is a pixelThe coded prediction value of (a) is,for the ith pixel value of the nth frame of the right view at the decoder end,for the ith pixel value in the n-1 th frame of the right view at the decoder side,for the error concealment value of the ith pixel of the (n-1) th frame of the right view at the decoder end,is the reconstructed value of the ith pixel in the nth frame of the right view at the encoder end,is a pixelThe probability of the authentication being successful is determined,is a pixelThe probability of the authentication being successful is determined,is a pixelThe residual values of the encoding of (1),is a pixelThe coded prediction value of (a) is,is a pixelA coding prediction value in a left view; the data can be directly obtained in the encoding process;andobtained by step 7;

9) if minD > D, minD is equal to D, and m _ opt is equal to m, MCS _ opt is equal to MCS, QP _ opt is equal to QP, and MCS is equal to MCS + 1;

10) if the MCS is less than or equal to maxMCS, returning to the step 6;

11) if m is less than or equal to maxH, returning to the step 4;

12) QP + 1; if the QP is less than or equal to maxQP, returning to the step 3;

13) and outputting m _ opt, MCS _ opt and QP _ opt, and using the parameters to encode the current 3D video data packet, configure the number of Hash authentication father nodes, select a physical layer MCS and perform wireless transmission.