CN108427828A - A kind of device of automatic assessment planar design placement quality and optimization - Google Patents

Abstract

The present invention relates to a kind of picture and text planar design technology fields, quality evaluation can be carried out to design finished product by providing one kind, to carry out self-service adjustment and optimization to design finished product, the device based on automatic assessment planar design placement quality and optimization of the quality of design finished product is improved.

Description

Device for automatically evaluating layout quality and optimizing planar design

Technical Field

The invention relates to the technical field of graphic and text planar design, in particular to a device for automatically evaluating the planar design layout quality and optimizing.

Background

As the advertising industry develops, people have more and more demands on the flat design. Despite the advent of many computer aided design tools (e.g., Photoshop, SketchUp, Illustrator, etc.), poster design is greatly facilitated, increasing the efficiency of the design. However, the way of realizing poster design by proprietary graphic design software such as Photoshop and CorelDRAW has high professional requirements, and for most people without plane design theory knowledge and basic aesthetic knowledge, it is still a difficult matter to do professional plane design.

Currently, some self-service flat design software realizes simple graphic and text flat design through limited functions. The self-service plane design method is mainly characterized in that a design template is provided for a user in advance, a user-defined area is reserved at a fixed position of the design template, and the user adds image-text materials in the user-defined area. Although the image-text planar design mode is simple to realize and low in professional requirement and can be suitable for the common masses, the final effect of the image-text planar design is limited by the inherent design template, and moreover, because the system lacks an objective judgment standard for the quality of a final design finished product, the design finished product cannot be effectively adjusted and optimized by self, the quality of the design finished product cannot be effectively improved by self, the personalized design requirement of a user cannot be met to a great extent, and the practicability is low.

Disclosure of Invention

Therefore, in order to solve the above problems, the present invention provides an apparatus for automatically evaluating the quality and optimization of a planar design layout, which is capable of evaluating the quality of a designed finished product, so as to perform self-help adjustment and optimization on the designed finished product, thereby improving the quality of the designed finished product.

In order to realize the technical problem, the solution scheme adopted by the invention is as follows: a device for automatically evaluating the quality and optimization of a planar design layout comprises a design structure analysis module, a grading module and an optimization design module;

the design structure analysis module converts a picture format file of a designed finished product into a structured file, and can obtain a text frame attribute and a background picture attribute of the designed finished product according to the structured file, wherein the text frame attribute comprises the frame size of a text frame and the position of the text frame, and the background picture attribute comprises the size of the background picture and color elements of the background picture;

the scoring module obtains the character frame attribute and the background image attribute through the structured file output by the structure analysis module so as to perform design quality scoring, wherein the design quality scoring comprises alignment feature evaluation on the character frame, regularity feature evaluation on the character frame, balance feature evaluation on the character frame and the image, blank feature evaluation on the poster and overlapping feature evaluation on the character frame and the image;

the optimization design module automatically carries out parameter change operation on each constituent parameter of the text box attribute and the background graph attribute respectively, and marks the planar design layout state obtained after the nth parameter change operation as an environmental state S_nAmbient state S_nCalculating the environment state S by superposing all the characteristic evaluation scores obtained by the scoring module_nThe environmental score function of (1), (Sn); meanwhile, the optimization design module calculates an action execution probability function P (an) of the nth parameter change operation according to the average distribution probability, wherein P (an) = 1/N, and N represents the total number of parameter change operations; the optimization design module obtains the nth parameter change according to an environment score function R (Sn) and an action execution probability function P (an)Action expectation value E of operation_n，E_n= p (an) x R (Sn); the optimization design module changes the action expectation value E corresponding to each parameter_nPerforming accumulation calculation to obtain the maximum expected action value E_maxAnd the action expectation value E is set_maxAnd (4) executing the corresponding parameter change operation on the design finished product to obtain the optimized design finished product.

Further, the alignment characteristics of the text boxes are evaluated to calculate the alignment of all the text boxes; the alignment characteristics for the text box were evaluated as:

① calculating the coordinate positions of 6 dimensions including upper, lower, left, right, horizontal middle point and vertical middle point of each frame, and marking each frame as T₁， T₂，T_3…T_n，T_nThe data structure representing the dimension coordinate position of the nth text box and the text box frame is represented as T_n{ T_n(up), T_n(down), T_n(left), T_n(right), T_n(horizion), T_n(veritical)}；

② calculating histograms of the 6 dimensional coordinate positions, calculating the average M of the 6 dimensional coordinate positions:

M_up= (T_1(up)+ T_2(up)+…+ T_n(up))/2；

M_down= (T_1(down)+ T_2(down)+ … + T_n(down))/2；

M_left= (T_1(left)+ T_2(left)+ … + T_n(left))/2；

M_right= (T_1(lright)+ T_2(lright)+ … + T_n(lright))/2；

M_horizion= (T_1(horizion)+ T_2(horizion)+ … + T_n(horizion))/2；

M_veritical= (T_1(veritical)+ T_2(veritical)+ … + T_n(veritical))/2；

③ calculating the discrete degree D of the 6 dimensional coordinate positions according to the histograms of the 6 dimensional coordinate positions, wherein the discrete degree is the alignment score;

；

；

；

；

；

；

④ selecting one dimension with lowest discrete degree from 6 dimension coordinate positions as one alignment mode, and selecting the discrete degree fraction of the alignment mode as the final alignment fraction S_{Alignment of}I.e. by

S_{Alignment of}= min(D_up, D_down,D_left, D_right, D_horizion, D_vertical)。

Further, the regularity characteristics of the text boxes are evaluated to calculate the width or height consistency of each text box frame; the regularity characteristics for the text box were evaluated as:

① obtaining the width and height data of each text box frame, marking each text box as T₁， T₂，T_3…T_n，T_nRepresenting the nth text box, the dimensional coordinate position data structure of the text box frame being represented as T_n{ T_n(width), T_n(hight)}；

② calculate the average M of the width and height data for each frame:

M_width= （T_1（width）+ T_2(width)+ …+ T_n(width)）/3；

M_hight= （T_1(hight)+ T_2(hight)+…+ T_n(hight)/3；

③ calculate the variance of the height and width data for each frame:

；

；

④ selecting the minimum D of the two variances_{min =}min(D_width，D_hight)；

⑤ the minimum variance is normalized to obtain a regularity score S_{Normalization}= 1/(1+D_min)。

Further, the balance characteristics of the text box and the picture are evaluated to calculate the upper-lower and left-right symmetry of the visual saliency map; the balance characteristics for the text box and the picture are evaluated as:

setting the importance of the text frame as 1, setting the importance of the background picture as an interval of 0-1, calculating a total graph after all the picture layers are superposed, and calculating a significance graph of the total graph:

① mark n text boxes as T₁，T₂，T_3…T_n， T₁=1, T₂=1, … T_n= 1; marking m background pictures as P respectively₁， P₂，P_3…P_m， P₁= random(0,1) = P₂= P₃= … = P_m；

② calculating text box saliency T_salAnd background Picture saliency P_sal：

N is the total number of text boxes, T_kRepresents the kth text box;

m is the total number of background pictures, P_kRepresenting the kth background picture;

③ calculating the significance P of the background picture_salSignificance with text box T_salOverlapping to obtain a final significance map P_fianl，I.e. P_fianl= P_sal+T_sal；

④ calculating the final saliency map P_fianlTo obtain a balance score S_Balancing：

，

Where m is the total number of background pictures,representing the significance of the left position of the kth significance map,representing the significance of the left position of the kth significance map,showing the significance of the upper edge position of the kth significance map,the saliency at the bottom position of the kth saliency map is shown.

Further, the blank characteristics of the poster are evaluated as the size of the poster area occupied by the text box and the picture saliency block, and the blank characteristics of the poster are evaluated as follows:

① calculating the margin ratio P of the poster content margin_white：

Area(T)_kArea of the K-th character frame, area (P)_kArea of the K background picture, area (total) of the design finished product;

② normalizing the blank ratio to obtain the evaluation score S of blank feature_{White space}：

。

Further, the overlapping characteristics of the text box and the picture are evaluated as:

①, calculating a total graph of all background pictures in the poster after being overlaid according to the image layers, and calculating a saliency map of the poster:

Area(P)_kthe area of the K background picture is shown, and m represents the total number of the background pictures;

② calculating the occlusion ratio:

wherein,the image area of the superposition position of the image area and the character area is represented;

③ normalization processing is carried out to the shielding proportion, and the overlapping characteristic score S is obtained after normalization processing_Overlap：

。

By adopting the technical scheme, the invention has the beneficial effects that: the device for automatically evaluating the layout quality and optimization of the planar design can convert aesthetic subjective standards of the design into measurable mathematical evaluation comparison, and further optimize a designed finished product through deep reinforcement learning according to the evaluation, thereby realizing self-help adjustment and optimization of the designed finished product and improving the quality of the designed finished product.

Drawings

FIG. 1 is a block diagram of the structural principles of an embodiment of the present invention.

Detailed Description

The invention will now be further described with reference to the accompanying drawings and specific embodiments.

Referring to fig. 1, an apparatus for automatically evaluating the quality and optimization of a planar design layout according to an embodiment of the present invention includes a design structure analysis module, a scoring module, and an optimization design module.

The design structure analysis module converts a picture format file of a designed finished product into a structured file, and can obtain a text frame attribute and a background picture attribute of the designed finished product according to the structured file, wherein the text frame attribute comprises the frame size of a text frame and the position of the text frame, and the background picture attribute comprises the size of the background picture and color elements of the background picture.

The scoring module obtains the character frame attribute and the background image attribute through the structured file output by the structure analysis module so as to carry out design quality scoring, wherein the design quality scoring comprises alignment feature evaluation on the character frame, regularity feature evaluation on the character frame, balance feature evaluation on the character frame and the image, blank feature evaluation on the poster, and overlapping feature evaluation on the character frame and the image:

evaluating the alignment characteristics of the text boxes to calculate the alignment of all the text boxes; the alignment characteristics for the text box were evaluated as:

① calculating the coordinate positions of 6 dimensions including upper, lower, left, right, horizontal middle point and vertical middle point of each frame, and marking each frame as T₁， T₂，T_3…T_n，T_nThe data structure representing the dimension coordinate position of the nth text box and the text box frame is represented as T_n{ T_n(up), T_n(down), T_n(left), T_n(right), T_n(horizion), T_n(veritical)}；

② calculating histograms of the 6 dimensional coordinate positions, calculating the average M of the 6 dimensional coordinate positions:

M_up= (T_1(up)+ T_2(up)+…+ T_n(up))/2；

M_down= (T_1(down)+ T_2(down)+ … + T_n(down))/2；

M_left= (T_1(left)+ T_2(left)+ … + T_n(left))/2；

M_right= (T_1(lright)+ T_2(lright)+ … + T_n(lright))/2；

M_horizion= (T_1(horizion)+ T_2(horizion)+ … + T_n(horizion))/2；

M_veritical= (T_1(veritical)+ T_2(veritical)+ … + T_n(veritical))/2；

③ calculating the discrete degree D of the 6 dimensional coordinate positions according to the histograms of the 6 dimensional coordinate positions, wherein the discrete degree is the alignment score;

；

；

；

；

；

；

④ selecting one dimension with lowest discrete degree from 6 dimension coordinate positions as one alignment mode, and selecting the discrete degree fraction of the alignment mode as the final alignment fraction S_{Alignment of}I.e. by

S_{Alignment of}= min(D_up, D_down,D_left, D_right, D_horizion, D_vertical)。

Evaluating the regularity characteristics of the text boxes to calculate the width or height consistency of each text box frame; the regularity characteristics for the text box were evaluated as:

① obtaining the width and height data of each text box frame, marking each text box as T₁， T₂，T_3…T_n，T_nRepresenting the nth text box, the dimensional coordinate position data structure of the text box frame being represented as T_n{ T_n(width), T_n(hight)}；

② calculate the average M of the width and height data for each frame:

M_width= （T_1（width）+ T_2(width)+ …+ T_n(width)）/3；

M_hight= （T_1(hight)+ T_2(hight)+…+ T_n(hight)/3；

③ calculate the variance of the height and width data for each frame:

；

；

④ selecting the minimum D of the two variances_{min =}min(D_width，D_hight)；

⑤ the minimum variance is normalized to obtain a regularity score S_{Normalization}= 1/(1+D_min)。

Evaluating the balance characteristics of the text box and the picture to calculate the upper-lower symmetry and the left-right symmetry of the visual saliency map; the balance characteristics for the text box and the picture are evaluated as:

setting the importance of the text frame as 1, setting the importance of the background picture as an interval of 0-1, calculating a total graph after all the picture layers are superposed, and calculating a significance graph of the total graph:

① mark n text boxes as T₁，T₂，T_3…T_n， T₁=1, T₂=1, … T_n= 1; marking m background pictures as P respectively₁， P₂，P_3…P_m， P₁= random(0,1) = P₂= P₃= … = P_m；

② calculating text box saliency T_salAnd background Picture saliency P_sal：

N is the total number of text boxes, T_kRepresents the kth text box;

m is the total number of background pictures, P_kRepresenting the kth background picture;

③ calculating the significance P of the background picture_salSignificance with text box T_salOverlapping to obtain a final significance map P_fianl，I.e. P_fianl= P_sal+T_sal；

④ calculating the final saliency map P_fianlTo obtain a balance score S_Balancing：

，

Where m is the total number of background pictures,representing the significance of the left position of the kth significance map,representing the significance of the left position of the kth significance map,showing the significance of the upper edge position of the kth significance map,the saliency at the bottom position of the kth saliency map is shown.

The blank features of the poster are evaluated as the size of the poster area occupied by the text box and the picture saliency block, and the blank features of the poster are evaluated as follows:

① calculating the margin ratio P of the poster content margin_white：

Area(T)_kArea of the K-th character frame, area (P)_kArea of the K background picture, area (total) of the design finished product;

② normalizing the blank ratio to obtain the evaluation score S of blank feature_{White space}：

。

The overlap characteristics for the text box and the picture are evaluated as:

①, calculating a total graph of all background pictures in the poster after being overlaid according to the image layers, and calculating a saliency map of the poster:

Area(P)_kthe area of the K background picture is shown, and m represents the total number of the background pictures;

② calculating the occlusion ratio:

wherein,the image area of the superposition position of the image area and the character area is represented;

③ normalization processing is carried out to the shielding proportion, and the overlapping characteristic score S is obtained after normalization processing_Overlap：

。

The optimization design module automatically carries out parameter change operation on each constituent parameter of the text box attribute and the background graph attribute respectively, and marks the planar design layout state obtained after the nth parameter change operation as an environmental state S_nAmbient state S_nObtained by superposing the characteristic evaluation scores obtained by the scoring module, namely S_n=S_{Alignment of}+S_{Normalization}+S_Balancing+S_{White space}+S_OverlapComputing environmental State S_nThe environmental score function of (1), (Sn); meanwhile, the optimization design module calculates an action execution probability function P (an) of the nth parameter change operation according to the average distribution probability, wherein P (an) = 1/N, and N represents the total number of parameter change operations; the optimization design module obtains an action expected value E of the nth parameter change operation according to an environment score function R (Sn) and an action execution probability function P (an)_n，E_n= p (an) x R (Sn); the optimization design module changes the action expectation value E corresponding to each parameter_nPerforming accumulation calculation to obtain the maximum expected action value E_maxAnd the action expectation value E is set_maxAnd (4) executing the corresponding parameter change operation on the design finished product to obtain the optimized design finished product.

In conclusion, the device for automatically evaluating the layout quality and optimization of the planar design can evaluate the quality of the designed finished product so as to adjust and optimize the designed finished product by self, thereby improving the quality of the designed finished product and having strong practicability.

The above description is only an embodiment utilizing the technical content of the present disclosure, and any modification and variation made by those skilled in the art can be covered by the claims of the present disclosure, and not limited to the embodiments disclosed.

Claims (6)

1. An automatic evaluate the planar design layout quality and optimization device, its characterized in that: the system comprises a design structure analysis module, a grading module and an optimization design module;

the design structure analysis module converts a picture format file of a designed finished product into a structured file, and can obtain a text frame attribute and a background picture attribute of the designed finished product according to the structured file, wherein the text frame attribute comprises the frame size of a text frame and the position of the text frame, and the background picture attribute comprises the size of the background picture and color elements of the background picture;

the scoring module obtains the character frame attribute and the background image attribute through the structured file output by the structure analysis module so as to perform design quality scoring, wherein the design quality scoring comprises alignment feature evaluation on the character frame, regularity feature evaluation on the character frame, balance feature evaluation on the character frame and the image, blank feature evaluation on the poster and overlapping feature evaluation on the character frame and the image;

the optimization design module automatically carries out parameter change operation on each constituent parameter of the text box attribute and the background graph attribute respectively, and marks the planar design layout state obtained after the nth parameter change operation as an environmental state S_nAmbient state S_nCalculating the environment state S by superposing all the characteristic evaluation scores obtained by the scoring module_nThe environmental score function of (1), (Sn); meanwhile, the optimization design module calculates an action execution probability function P (an) of the nth parameter change operation according to the average distribution probability, wherein P (an) = 1/N, and N represents the total number of parameter change operations; the optimization design module obtains an action expected value E of the nth parameter change operation according to an environment score function R (Sn) and an action execution probability function P (an)_n，E_n= p (an) x R (Sn); the optimization design module changes the action expectation value E corresponding to each parameter_nPerforming accumulation calculation to obtain the maximum expected action value E_maxAnd the action expectation value E is set_maxAnd (4) executing the corresponding parameter change operation on the design finished product to obtain the optimized design finished product.

2. The apparatus for automatically evaluating the quality and optimization of a floor plan layout of claim 1 wherein the alignment characteristics of the text boxes are evaluated to calculate the alignment of all text boxes; the alignment characteristics for the text box were evaluated as:

① calculating the coordinate positions of 6 dimensions including upper, lower, left, right, horizontal middle point and vertical middle point of each frame, and marking each frame as T₁， T₂，T_3…T_n，T_nDenotes the n-thThe data structure of the dimension coordinate position of the individual text box and the text box frame is represented as T_n{ T_n(up), T_n(down), T_n(left), T_n(right), T_n(horizion), T_n(veritical)}；

② calculating histograms of the 6 dimensional coordinate positions, calculating the average M of the 6 dimensional coordinate positions:

M_up= (T_1(up)+ T_2(up)+…+ T_n(up))/2；

M_down= (T_1(down)+ T_2(down)+ … + T_n(down))/2；

M_left= (T_1(left)+ T_2(left)+ … + T_n(left))/2；

M_right= (T_1(lright)+ T_2(lright)+ … + T_n(lright))/2；

M_horizion= (T_1(horizion)+ T_2(horizion)+ … + T_n(horizion))/2；

M_veritical= (T_1(veritical)+ T_2(veritical)+ … + T_n(veritical))/2；

③ calculating the discrete degree D of the 6 dimensional coordinate positions according to the histograms of the 6 dimensional coordinate positions, wherein the discrete degree is the alignment score;

；

；

；

；

；

；

④ selecting one dimension with lowest discrete degree from 6 dimension coordinate positions as one alignment mode, and selecting the discrete degree fraction of the alignment mode as the final alignment fraction S_{Alignment of}I.e. by

S_{Alignment of}= min(D_up, D_down,D_left, D_right, D_horizion, D_vertical)。

3. The apparatus for automatically evaluating layout quality and optimization of a floor plan as claimed in claim 1, wherein the regularity characteristic of the text box is evaluated to calculate the consistency of the width or height of each text box frame; the regularity characteristics for the text box were evaluated as:

① obtaining the width and height data of each text box frame, marking each text box as T₁， T₂，T_3…T_n，T_nRepresenting the nth text box, the dimensional coordinate position data structure of the text box frame being represented as T_n{ T_n(width), T_n(hight)}；

② calculate the average M of the width and height data for each frame:

M_width= （T_1（width）+ T_2(width)+ …+ T_n(width)）/3；

M_hight= （T_1(hight)+ T_2(hight)+…+ T_n(hight)/3；

③ calculate the variance of the height and width data for each frame:

；

；

④ selecting the minimum D of the two variances_{min =}min(D_width，D_hight)；

⑤ the minimum variance is normalized to obtain a regularity score S_{Normalization}= 1/(1+D_min)。

4. The apparatus for automatically evaluating the quality and optimization of a floor plan layout according to claim 1, wherein the balance features of the text box and the picture are evaluated to calculate the upper and lower and left and right symmetries of the visual saliency map; the balance characteristics for the text box and the picture are evaluated as:

setting the importance of the text frame as 1, setting the importance of the background picture as an interval of 0-1, calculating a total graph after all the picture layers are superposed, and calculating a significance graph of the total graph:

① mark n text boxes as T₁，T₂，T_3…T_n， T₁=1, T₂=1, … T_n= 1; marking m background pictures as P respectively₁， P₂，P_3…P_m， P₁= random(0,1) = P₂= P₃= … = P_m；

② calculating text box saliency T_salAnd background Picture saliency P_sal：

N is the total number of text boxes, T_kRepresents the kth text box;

m is the total number of background pictures, P_kRepresenting the kth background picture;

③ calculating the significance P of the background picture_salSignificance with text box T_salOverlapping to obtain a final significance map P_fianl，I.e. P_fianl= P_sal+T_sal；

④ calculating the final saliency map P_fianlTo obtain a balance score S_Balancing：

，

Where m is the total number of background pictures,representing the significance of the left position of the kth significance map,representing the significance of the left position of the kth significance map,showing the significance of the upper edge position of the kth significance map,the saliency at the bottom position of the kth saliency map is shown.

5. The apparatus for automatically evaluating the quality and optimization of a floor plan layout of claim 1 wherein the whiteout characteristics of the poster are evaluated as occupying the size of the poster area for the text box and the graphic saliency blocks, the whiteout characteristics of the poster are evaluated as:

① calculating the margin ratio P of the poster content margin_white：

Area(T)_kArea of the K-th letter box, Area(P)_kArea of the K background picture, area (total) of the design finished product;

② normalizing the blank ratio to obtain the evaluation score S of blank feature_{White space}：

。

6. The apparatus for automatically evaluating the quality and optimization of a floor plan layout as claimed in claim 1, wherein the overlap characteristics of the text box and the picture are evaluated as:

①, calculating a total graph of all background pictures in the poster after being overlaid according to the image layers, and calculating a saliency map of the poster:

Area(P)_kthe area of the K background picture is shown, and m represents the total number of the background pictures;

② calculating the occlusion ratio:

wherein,the image area of the superposition position of the image area and the character area is represented;

③ normalization processing is carried out to the shielding proportion, and the overlapping characteristic score S is obtained after normalization processing_Overlap：

。