Temporal Interface Parameters

The temporal information is used at each level for image processing. The time domain of level 1 can be processed based on layers (layer 0 and layer 1). There are two sets of interfaces in each time domain corresponding to two layers. If the interface has multiple parameters, interfaces with suffixes 0 and 1 are used to differentiate layers, for example, nXtfr0 and nXtfr1. For IPC application scenarios, layer-based processing is recommended. Layer 1 is set as the background layer, while layer 0 is set as the foreground layer. The other three levels adopt the single-layer structure. Level 3 uses the spatial processing mode with the temporal auxiliary.

• ref: determines whether to use the reference frame. When ref of level 0 is set to 0, only reference frame at level 0 is unavailable. ref of level 1 is the global switch of the temporal reference. When it is set to 0, temporal filtering is invalid and the reference frame is not loaded. When it is set to 1, temporal filtering takes effect. The recommended value is 1 for IPCs.
• tedge: indicates whether to process the regions with possible ghosting. The value 0 means to disable this function. The value 1 means to process the possible ghosting. The values 2 and 3 mean to process the possible blurring. For IPCs, the recommended value is 0 and should not be changed.
• nXstr: indicates the spatial-temporal filtering strength for NR. Certain mask noise may be introduced. A larger value indicates better NR effect but a higher probability of introducing mask noise. The value range is [0, 31].
• nXsdz: cooperates with the nXstr interface to limit the spatial filter. The value range is [0, 999]. A smaller value indicates a higher nXstr strength. The value 999 means to disable the spatial filter of this level.
• nXtss: A larger value indicates smoother static regions but more blurs in static regions of the image. The value range is [0, 15]. The two values of this parameter represent different regions.
  • If the processing is not region-based, you can adjust this parameter to make the background smooth.
  • If the processing is region-based, you can increase the first parameter to prevent foreground blurring in regions where the motion is not obvious. The second parameter applies to the background.
• nXtsi: cooperates with nXtss. The value can be 0 or 1, which indicates different processing modes. The value 0 indicates the pure spatial processing mode, and the value 1 indicates the spatial-temporal processing mode. For details, see the value of mXid.
• nXtfs: specifies the temporal filter strength. When the current filter region uses temporal filter, this parameter indicates the strength of the temporal filtering effects. A larger value indicates a greater strength. The value range is [0, 15].
• nXtfr: controls the balance between ghosting and NR. There are six processing modes. A smaller value indicates stronger control over ghosting but a weaker NR capability. The mode with the optimal NR effects is used. The value range is [0, 31].
• nXtfrs: works together with nXtfr, valid only at level 0. A larger value indicates stronger spatial effect.
• nXdzm: specifies the selection mode of tdz. The value range is [0, 1].
• [nXtdz]: protects the texture or enhances the denoising effect. The value range is [0, 999].
  • When dzm is set to 0, the texture noise of the motion region is obvious but the texture does not disappear or get blurred. When tdz is set to a large value, the texture of the motion region can be protected, but the NR effect decreases.
  • When dzm is set to 1, the temporal filtering strength of motion regions is increased. When tdz is set to a larger value, the temporal NR strength of motion regions is enhanced.
• nXtdx: Indicates the parameter function strength of nXtdz. The value range is [0, 3]. A smaller value indicates the stronger nXtdz parameter function. You are advised to retain the default value 2.
• mXmath/mXmathd: indicates the upper and lower limits of the motion detection threshold. A larger value indicates that more pixels are judged as static by the motion detection unit, temporal filtering is performed on more pixels, and the image noise is less obvious. Generally, you are advised to set tfs to the maximum value and adjust mXmath/mXmathd to suppress raindrops. Then, decrease the tfs value until no raindrop appears. If layer-based processing is used, the system divides the static region of the image based on the second value of the interface and uses it as the background layer (that is, absolute static region). The remaining image is used as the foreground layer. The relative static region and the motion region are further divided according to the first parameter, and are processed separately. Level-3 mXmath/mXmathd is used to divide images into motion and static regions, so that the regions can be filtered or enhanced separately. However, temporal filtering is unavailable. The value range is [0, 999]. mXmath must be greater than or equal to mXmathd.
• mXid: specifies the output effect based on the regions distinguished according to the results of mXmath. The value range of each parameter is [0, 3], which indicates the output result of sfr, nXstr, nXtfr, and nXtfs, respectively. A larger value indicates a stronger effect of the temporal parameter.
  • The first parameter indicates the processing mode chosen for the image content judged to belong to the motion region (where the feature variance is greater than or equal to the math value). You are advised to set this parameter to 0 or 1.
  • The third parameter indicates the processing mode chosen for the image content judged to belong to the static region (where the feature variance is less than the math value). You are advised to set this parameter to 2 or 3.
  • The second parameter takes effect only when madz and mabr are valid. The value range is [0, 3]. By default, the second parameter does not take effect. The effect is equivalent to performing extra processing in the static region based on madz and mabr after motion detection.
  • For layer-based processing at level 1, you do not need to set mXid1 for the background layer. You need to set only the three parameters of mXid0 for the foreground layer.
• AdvMath: specifies the common motion judgment interface math or the enhanced motion judgment interface to use. It is recommended that this switch be enabled when layer-based processing is used. The enhanced mode applies only to the foreground layer of level 1. When the enhanced interface is used, the value of math at the foreground layer is usually smaller than that of the common interface. The value range is [0, 1]. 0 indicates the common mode, and 1 indicates the enhanced mode. If the processing is not region-based, the value of this interface is 0.
• AdvTh: determines the strength of the enhanced math. A smaller value indicates greater strength.
• mXmabr and mXmadz: filter selection corresponds to the second parameter of mXid.
  • mXmabr: controls extra processing based on the luminance. A larger value indicates that more content is considered as the extra processing region, while less content is considered as the static region. The value range of each interface parameter is [0, 255].
  • mXmadz: controls extra processing based on image features. A smaller value indicates that more content is considered as the static region. A larger value indicates that more content is considered as the extra processing region. The value range is [0, 511].

  [Note] When madz or mabr is set to 0, the functions of the two interfaces are disabled. For IPC application scenarios, it is recommended that mabr and madz be set to 0 and not changed.

• mXmate: indicates the motion detection index of flat regions. A larger value indicates that more pixels are judged as static by the motion detection unit, temporal filtering is performed on more pixels, and the image noise is less obvious. Generally, you are advised to set math to a proper value, and then slightly set mate to balance raindrop noise and motion smearing. The value range is [0, 8].
• mXmatw: indicates the smearing prevention index of temporal filtering. A larger value indicates faster motion smearing convergence, and vice versa. The value range is [0, 3]. You are advised to set this parameter to the default value 2 and not change it.
• mXmabw: specifies the window size of the motion detection content. This parameter is used in conjunction with math. A larger value indicates a larger window. In low illuminance, if the raindrop phenomenon cannot be suppressed even when math is set to a large value, you are advised to set the mabw value to greater than 7, thereby relieving the burden of math to suppress raindrops and reducing side effects of the time domain. The value range is [0, 9]. If layer-based processing is used at a level, the value range of mabw of the background layer (layer 1) is [5, 9], and the value range of the foreground layer (layer 0) is [0, 9]. However, [0, 4] is recommended to prevent smearing. If layer-based processing is not used (that is, biPath = 0), the value range of mabw is [5, 9].
• mXmasw: indicates the raindrop prevention index of temporal filtering. A larger value helps reduce the probability of raindrop noise. The value range is [0,15]. The default value is 12, which should not be changed.