# 3x3 Convolution

This operator applies a user-specified 3x3 convolution matrix to the echogram.

You define the convolution matrix on the 3x3 Convolution page of the Variable Properties dialog box.

Echoview accepts a single operand of the following data types as input:

- Color
- Linear
- Power dB
- Sv
- TS
- Unspecified dB

**Note:** Under live viewing, the Live Export based on a convolution variable may be limited.

## Settings

The 3x3 Convolution Variable Properties dialog box pages include (common) Variable Properties pages and these operator pages:

### Operands page

### 3x3 Convolution page

Specify the 9 values of the kernel for the 3x3 convolution matrix.

## Algorithm

A convolution is effectively an integration of two functions. One function is the convolution kernel and the other is the signal data. In Echoview, this is accomplished by using this operator's ** custom convolution kernel and echogram sample data**. First, corresponding cells for the kernel and echogram sample space are multiplied. Then the values are summed and the result is stored in the corresponding output cell – for the 3x3 kernel this is the center cell.

A convolution custom kernel can be used for feature filtering in echogram data. The size and choice of kernel cell values is important in the composition of a feature filter. The 3x3 kernel allows the convolution to operate on a central sample and its nearest neighbors. Larger kernel filters (5x5 and 7x7) can deal with more distant neighbors or preferential directions in the data.

Filter types include blurring (for noise handling), sharpening, embossing, edge detection etc. A useful tip is to have the sum of kernel coefficients sum to 1 if they are positive (as in the blur kernel), or be between 0 and 1, to maintain an image balance between the values and not end up with output values that are all black (darker), or all white (brighter).

Refer to example kernels and output on Wikipedia Kernel (image processing) (2023) [Online] available at https://en.wikipedia.org/wiki/Kernel_(image_processing)

For details on how convolutions are applied in Echoview see Convolution algorithms: Sliding window and kernel.