DMPAC Dense Optical Flow Kernels.
|
void | tivxRegisterHwaTargetDmpacDofKernels (void) |
| Function to register HWA Kernels on the dmpac_dof Target.
|
|
void | tivxUnRegisterHwaTargetDmpacDofKernels (void) |
| Function to un-register HWA Kernels on the dmpac_dof Target.
|
|
void | tivxRegisterHwaTargetArmKernels (void) |
| Function to register HWA Kernels on the arm Target.
|
|
void | tivxUnRegisterHwaTargetArmKernels (void) |
| Function to un-register HWA Kernels on the arm Target.
|
|
VX_API_ENTRY vx_node VX_API_CALL | tivxDmpacDofNode (vx_graph graph, vx_user_data_object configuration, vx_image input_current_base, vx_image input_reference_base, vx_pyramid input_current, vx_pyramid input_reference, vx_image flow_vector_in, vx_user_data_object sparse_of_config, vx_image sparse_of_map, vx_image flow_vector_out, vx_distribution confidence_histogram) |
| [Graph] Creates a DMPAC_DOF Node. More...
|
|
VX_API_ENTRY vx_node VX_API_CALL | tivxDofVisualizeNode (vx_graph graph, vx_image flow_vector, vx_scalar confidence_threshold, vx_image flow_vector_rgb, vx_image confidence_image) |
| [Graph] Creates a DOF visualization node Node. More...
|
|
void | tivx_dmpac_dof_params_init (tivx_dmpac_dof_params_t *prms) |
| Function to initialize DOF parameters with default value. More...
|
|
void | tivx_dmpac_dof_sof_params_init (tivx_dmpac_dof_sof_params_t *prms) |
| Function to initialize DOF SOF parameters with default value. More...
|
|
void | tivx_dmpac_dof_cs_tree_params_init (tivx_dmpac_dof_cs_tree_params_t *prms) |
| Function to initialize DOF CS tree parameters with default value. More...
|
|
void | tivx_dmpac_dof_hts_bw_limit_params_init (tivx_dmpac_dof_hts_bw_limit_params_t *prms) |
| Function to initialize DOF HTS bandwidth limit parameters with default value. More...
|
|
◆ tivxDmpacDofNode()
VX_API_ENTRY vx_node VX_API_CALL tivxDmpacDofNode |
( |
vx_graph |
graph, |
|
|
vx_user_data_object |
configuration, |
|
|
vx_image |
input_current_base, |
|
|
vx_image |
input_reference_base, |
|
|
vx_pyramid |
input_current, |
|
|
vx_pyramid |
input_reference, |
|
|
vx_image |
flow_vector_in, |
|
|
vx_user_data_object |
sparse_of_config, |
|
|
vx_image |
sparse_of_map, |
|
|
vx_image |
flow_vector_out, |
|
|
vx_distribution |
confidence_histogram |
|
) |
| |
[Graph] Creates a DMPAC_DOF Node.
- The data format of image within pyramid MUST be
VX_DF_IMAGE_U8
, VX_DF_IMAGE_U16
, or TIVX_DF_IMAGE_P12
format.
- The max size of the base image of pyramid is 2048 width, and 1024 height.
- The pyramid MUST use scale of VX_SCALE_PYRAMID_HALF
- The max number of pyramid levels can be 6
- The width and height of base level MUST be interger multiple of 2^pyramidlevels
- The meta properties of input_current, input_reference MUST be identical
- If the optional input_current_base and input_reference_base is used, then the base size of the respective pyramid paramters should be half the width and height of the base images.
- Each flow vector sample can be encoded in
VX_DF_IMAGE_U32
format or VX_DF_IMAGE_U16
format. Note that the VX_DF_IMAGE_U32
format is recommended since it contains additional fractional bits of the flow vectors as well as confidence score information, while the VX_DF_IMAGE_U16
format only contains integer flow vector information. Also, because of this, the temporal predictor can only be used with the VX_DF_IMAGE_U32
format.
- 32bit encoding format (u,v, confidence score) is as follows:
- Confidence (4 bit)
- [3:0] Confidence is 4 bits (16 levels of confidence value)
- Vertical flow vector (7bit signed integer, 4 bit fractional):
- [7:4] Fractional is 4 bits (support 1/16th pixel of precision)
- [14:8] Signed Integer is 7 bits (support up to +63 to -63 pixel Vertical flow vectors)
- [15] Copy of signed bit from integer
- Horizontal flow vector (9bit signed integer, 4 bit fractional):
- [19:16] Fractional is 4 bits (support 1/16th pixel of precision)
- [28:20] Signed Integer is 9 bits (support up to +255 to -255 pixel Horizontal flow vectors)
- [31:29] Copies of signed bit from integer
- 16bit encoding format (u,v) is as follows:
- Vertical flow vector (7bit signed integer):
- [6:0] Signed Integer is 7 bits (support up to +63 to -63 pixel Vertical flow vectors)
- Horizontal flow vector (9bit signed integer):
- [15:7] Signed Integer is 9 bits (support up to +255 to -255 pixel Horizontal flow vectors)
- Parameters
-
[in] | graph | The reference to the graph. |
[in] | configuration | The input object of a single params structure of type tivx_dmpac_dof_params_t . |
[in] | input_current_base | (optional) Current input base image of pyramid (if not included in input_current pyramid). |
[in] | input_reference_base | (optional) Reference input base image of pyramid (if not included in input_reference pyramid). |
[in] | input_current | Current input pyramid. |
[in] | input_reference | Reference input pyramid. |
[in] | flow_vector_in | (optional) Flow vector from previous execution of DOF. Size of image is base_width x base_height x size of data format. Use VX_DF_IMAGE_U32 dataformat. |
[in] | sparse_of_config | (optional) The input object of a single params structure of type tivx_dmpac_dof_sof_params_t . This can change from frame to frame and should be syncronized with changes to the sparse_of_map input. |
[in] | sparse_of_map | (optional) Sparse OF bit-mask of the input image. Size of image is base_width/8 x base_height. Use VX_DF_IMAGE_U8 dataformat. |
[out] | flow_vector_out | Flow vector output. When sparse_of_map is disabled, size of image is base_width x base_height x size of data format. Use VX_DF_IMAGE_U32 or VX_DF_IMAGE_U16 dataformat. VX_DF_IMAGE_U32 format include confidence score and subpel fractional flow vectors, while VX_DF_IMAGE_U16 is only integer flow vectors. |
[out] | confidence_histogram | (optional) Confidence histogram. Distribution meta properties, num_bins = 16, offset = 0, range = 16. |
- See also
TIVX_KERNEL_DMPAC_DOF_NAME
- Returns
vx_node
.
- Return values
-
vx_node | A node reference. Any possible errors preventing a successful creation should be checked using vxGetStatus |
◆ tivxDofVisualizeNode()
[Graph] Creates a DOF visualization node Node.
- Parameters
-
[in] | graph | The reference to the graph. |
[in] | flow_vector | Flow vector output from dmpac_dof node |
[in] | confidence_threshold | (optional) Threshold to use when generating flow_vector_rgb. vx_scalar of type vx_uint32. Valid values are 0 (low threshold/confidence) .. 15 (high threshold/confidence). When NULL, default value of 8 is used. |
[out] | flow_vector_rgb | flow vector representated as 24 RGB image |
[out] | confidence_image | confidence values represented as U8 grayscale image, 255 is high confidence |
- See also
TIVX_KERNEL_DOF_VISUALIZE_NAME
- Returns
vx_node
.
- Return values
-
vx_node | A node reference. Any possible errors preventing a successful creation should be checked using vxGetStatus |
◆ tivx_dmpac_dof_params_init()
Function to initialize DOF parameters with default value.
- Parameters
-
prms | [IN] Pointer to DOF parameter structure |
◆ tivx_dmpac_dof_sof_params_init()
Function to initialize DOF SOF parameters with default value.
- Parameters
-
prms | [IN] Pointer to DOF SOF parameter structure |
◆ tivx_dmpac_dof_cs_tree_params_init()
Function to initialize DOF CS tree parameters with default value.
- Parameters
-
prms | [IN] Pointer to DOF confidence score parameter structure |
◆ tivx_dmpac_dof_hts_bw_limit_params_init()
Function to initialize DOF HTS bandwidth limit parameters with default value.
- Parameters
-
prms | [IN] Pointer to DOF bandwidth limiter parameter structure |