Documentation

JPEGE

The JPEG encoder (JPEGE) encodes YUV images into JPG images. For details about the JPEGE functions and restrictions, see Functions and Restrictions.

This section describes the API call sequence of JPEGE, and sample code is also provided to help you better understand the sequence.

API Call Sequence

If a YUV image needs to be encoded into a JPEG image during app development, ensure that your app contains the code logic for such image encoding. For details about the API call sequence, see pyACL API Call Sequence.

Figure 1 JPEG image encoding

The current system supports YUV to JPG image encoding. The key APIs are described as follows:

  1. Call acl.media.dvpp_create_channel_desc to create a channel description.
  2. Call acl.media.dvpp_create_channel to create a channel for image data processing.
  3. Call acl.media.dvpp_create_jpege_config to create image encoding configuration data.
  4. Before encoding JPEG images, call acl.media.dvpp_malloc to allocate device buffers for storing the input or output data.

    Before allocating an output buffer, you can call acl.media.dvpp_jpeg_predict_enc_size to predict the output allocation size required for the encoded image based on the input image description and image encoding configuration data.

  5. Call the asynchronous API acl.media.dvpp_jpeg_encode_async for encoding.

    Also call acl.rt.synchronize_stream to block the application until all tasks in the specified stream are complete.

  6. Call acl.media.dvpp_destroy_jpege_config to destroy the image encoding configuration data.
  7. After the encoding is complete, call acl.media.dvpp_free to free the input and output buffers in a timely manner.
  8. Call acl.media.dvpp_destroy_channel to destroy the image data processing channel.

    Call acl.media.dvpp_destroy_channel_desc to destroy the channel description after the channel is destroyed.

Sample Code

You can view the complete code in Sample Obtaining.

After APIs are called, add an exception handling branch, and record error logs and warning logs. The following is a code snippet of key steps only, which is not ready to use.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
 
import acl
# ......

# 1. Initialize pyACL.
ret = acl.init()

# 2. Allocate runtime resources.

# 3. Create a description of the data processing channel. self.dvpp_channel_desc is of type acldvppChannelDesc.
self.dvpp_channel_desc = acl.media.dvpp_create_channel_desc()

# 4. Create a data processing channel.
ret = acl.media.dvpp_create_channel(self.dvpp_channel_desc)

# 5. Allocate buffers.
# 5.1 Input buffers.
# Load the YUV image data. Copy image data from the host to the device by calling acl.rt.memcpy.
np_jpg = np.fromfile(self.path, dtype=np.byte)
self.in_buffer_size = np_jpg.itemsize * np_jpg.size
bytes_data = np_jpg.tobytes()
np_jpg_ptr = acl.util.bytes_to_ptr(bytes_data)
self.in_buffer_dev, ret = acl.media.dvpp_malloc(self.in_buffer_size)
ret = acl.rt.memcpy(self.in_buffer_dev, self.in_buffer_size, np_jpg_ptr,
                    self.in_buffer_size, 1)

# 5.2 Calculate and allocate the output device buffer self.out_buffer_dev in 16-pixel-aligned mode to store the encoded output data.
height_stride = ((h + 1) // 2) * 2
width_stride = ((w + 15) // 16) * 16
# malloc for output
self.out_buffer_size = (width_stride * height_stride * 3) // 2
self.out_buffer_dev, ret = acl.media.dvpp_malloc(self.out_buffer_size

 # 6. Create the description of the input image and set the attribute values.
# JPEG encoding: Sets the input YUV image data and size to the image description.
self.pic_desc = acl.media.dvpp_create_pic_desc()
ret = acl.media.dvpp_set_pic_desc_data(self.pic_desc, self.in_buffer_dev)
ret = acl.media.dvpp_set_pic_desc_size(self.pic_desc, self.in_buffer_size)
ret = acl.media.dvpp_set_pic_desc_format(self.pic_desc, PIXEL_FORMAT_YUV_SEMIPLANAR_420)
ret = acl.media.dvpp_set_pic_desc_width(self.pic_desc, w)
ret = acl.media.dvpp_set_pic_desc_height(self.pic_desc, h)
ret = acl.media.dvpp_set_pic_desc_width_stride(self.pic_desc, width_stride)
ret = acl.media.dvpp_set_pic_desc_height_stride(self.pic_desc, height_stride)

# 7. Create image encoding configuration data, set the encoding quality, and predict the output image size, and allocate a buffer.
# Encoding quality in a range of [0, 100]. The encoding quality of level 0 is similar to that of level 100. For the value range of [1, 100], a smaller value indicates poorer output image quality.
self.jpege_config = acl.media.dvpp_create_jpege_config()
ret = acl.media.dvpp_set_jpege_config_level(self.jpege_config, 100)
self.out_buffer_size, ret = acl.media.dvpp_jpeg_predict_enc_size(
         self.pic_desc, self.jpege_config)
self.out_buffer_dev, ret = acl.media.dvpp_malloc(self.out_buffer_size)

# 8. Perform asynchronous encoding and call acl.rt.synchronize_stream to block the host until all tasks in the specified stream are complete.
acl.rt.set_context(self.context)
np_out_size = np.array([self.out_buffer_size], dtype=np.int32)
bytes_data = np_out_size.tobytes()
np_out_size_ptr = acl.util.bytes_to_ptr(bytes_data)
ret = acl.media.dvpp_jpeg_encode_async(self.dvpp_channel_desc, self.pic_desc,
                                       self.out_buffer_dev, np_out_size_ptr,
                                       self.jpege_config, self.stream)
ret = acl.rt.synchronize_stream(self.stream)

# 9. Allocate host memory hostPtr.
# 9.1 Load the encoded output image to the host, write the host memory data to the file, and free the host memory by calling acl.rt.free_host in a timely manner.
# 9.2 Create a NumPy object and convert it into data on the host to store the output.
np_output = np.zeros(size, dtype=np.byte)
bytes_data = np_output.tobytes()
np_output_ptr = acl.util.bytes_to_ptr(bytes_data)
size = int(np_out_size[0])
ret = acl.rt.memcpy(np_output_ptr, size, self.out_buffer_dev, size, 2)

# 10. After the decoding is complete, destroy the allocations, including the descriptions of the input and output images, input and output buffers, description of the channel as well as the channel.
ret = acl.media.dvpp_destroy_jpege_config(self.jpege_config)
ret = acl.media.dvpp_destroy_pic_desc(self.pic_desc)
ret = acl.media.dvpp_free(self.in_buffer_dev)
ret = acl.media.dvpp_free(self.out_buffer_dev)
ret = acl.media.dvpp_destroy_channel(self.dvpp_channel_desc)
ret = acl.media.dvpp_destroy_channel_desc(self.dvpp_channel_desc)

# 11. Destroy runtime allocations.
# Deinitialize pyACL.
ret = acl.finalize()
# ....
Parent topic: Media Data Processing V1